The Chesapeake Bay and the Control of
NOx Emissions: A Policy Analysis
Alan Krupnick, Virginia McConnell, David
Austin, Matt Cannon, Terrell Stoessell, and
Brian Morton
Discussion Paper 98-46
August 1998
1616 P Street, NW
Washington, DC 20036
Telephone 202-328-5000
Fax 202-939-3460
© 1998 Resources for the Future. All rights reserved.
No portion of this paper may be reproduced without
permission of the authors.
Discussion papers are research materials circulated by their
authors for purposes of information and discussion. They
have not undergone formal peer review or the editorial
treatment accorded RFF books and other publications.
The Chesapeake Bay and the Control of NOx Emissions:
A Policy Analysis
Alan Krupnick, Virginia McConnell, David Austin,
Matt Cannon, Terrell Stoessell, and Brian Morton
Abstract
Nitrogen oxide emissions not only affect air quality but have recently been found to be an important
source of nitrate pollution in the Chesapeake Bay. This analysis examines the costs, emissions, sourcespecific and location-specific allocations of NOx emissions reductions and the ancillary ozone related health
benefits under a range of policy scenarios. The paper includes analysis of three separate policies. The first
is a detailed analysis of the effect on nitrate loadings to the Bay of command and control policies specified
in the Clean Air Act and as part of the OTAG process. The second is a comparison of alternative scenarios
for reducing NOx emissions that meet nitrate loading goals, with or without concern for reducing ozone
concentrations and the health effects they cause. The third is a comparison of alternative approaches to
allocate NOx emissions to meet NOx reduction and ozone exposure goals while capturing the ancillary
effect on nitrate loadings. This last analysis focuses on the stake the Bay jurisdictions have in the outcome
of negotiations over NOx trading programs being developed by EPA for reducing ozone in the Eastern U.S.
With the primary focus on the Chesapeake Bay jurisdiction, all three analyses integrate the ancillary ozone
benefits of policies to reduce nitrate pollution, including examination of how these ancillary benefits change
under alternative meteorological episodes, and explore lower cost alternatives to current regulatory
programs in both qualitative and quantitative terms.
We find that the Chesapeake Bay benefits from efforts to reduce NOx emissions to meet the
ambient air quality standard for ozone. Airborne NOx emission reductions slated to occur under the Clean
Air Act in the Bay airshed will reduce nitrate loadings to the Bay by about 27 percent of the baseline
airborne levels. The additional controls of NOx contemplated in what we term the OTAG scenario is
estimated to result in an additional 20 percent reduction from this baseline. However, the paper's analysis
of possible least cost options shows that the costs of obtaining such reductions can be significantly reduced
by rearranging the allocation of emissions reductions to take advantage of source-type and locational
considerations. In addition, we find that adding consideration of ancillary ozone-related health benefits to
the picture does not alter any qualitative conclusions. Quantitatively, unless a link between ozone and
mortality risk is assumed, the benefits are too small to affect the cost-saving allocations of NOx reductions.
If the case for such a link can be made, the results change dramatically, with large overall increases in
NOx reductions and a relative shift in controls to non-Bay states and utility sources. These specific effects
are sensitive to the source-receptor coefficients linking NOx to ozone, however.
Our analyses also suggest that the Bay jurisdictions have a stake in the outcome of the NOx trading
debate -- that some trading designs can lead to better outcomes for these jurisdictions than others.
Nevertheless, a common feature of cost-savings policies is that they both rearrange emissions reductions and,
in the aggregate, reduce emissions less than a command and control system. Thus, some trading regimes result
in significantly smaller loadings reductions (up to 25 percent smaller) than the command and control approach.
Key Words: Chesapeake Bay, cost effectiveness, air pollution
JEL Classification Numbers: Q20, Q25, Q28
ii
Acknowledgments
The authors would like to acknowledge Erica Laich and her staff at E. H. Pechan
Associates for providing the data on abatement options, cost and emissions for this project.
We would also like to thank Paul Guthrie and his staff at SAI for their work in providing
source-receptor coefficients linking NOx emissions to ozone and to Paul for his sound
judgment and counsel on the numerous assumptions undergirding an analysis like ours.
Within RFF, we had excellent initial research assistance from Gar Ragland and invaluable
discussions with Dallas Burtraw and Karen Palmer. We would like to acknowledge our
sponsors -- EPA's Office of Policy, Planning, and Evaluation, particularly Willard Smith and
also Bob Noland, as well as EPA's Chesapeake Bay Program Office's Air Subcommittee and
the EPA's Acid Rain Office. Will Smith provided enormous moral, logistical, and intellectual
support as well. Moral and intellectual support also came from the Chesapeake Bay Program
Office, particularly Rick Batiuk and Julie Thomas. Finally, we would like to thank the
attendees at our two project briefings for their interest and comments -- many of those
mentioned above and other EPA staff as well as John Sherwell from Maryland's Department
of Natural Resources and Mike Uhart from NOAA.
iii
Table of Contents
Acknowledgments ........................................................................................................................... iii
I.
Introduction ........................................................................................................................... 1
II.
Previous Research .................................................................................................................. 3
III. Domains and Scenario Definitions .......................................................................................... 5
A.
Domains ....................................................................................................................... 5
B.
Scenarios ...................................................................................................................... 7
IV. The Conceptual Model ......................................................................................................... 10
V.
Data ..................................................................................................................................... 13
A.
Emissions and Costs ................................................................................................... 13
B.
NOx to Nitrate Source-Receptor Coefficients ............................................................. 14
C.
NOx to Ozone Source-Receptor Coefficients .............................................................. 15
D.
Ozone Benefits ........................................................................................................... 24
VI. Results of the Policy Analysis .............................................................................................. 26
Policy Analysis 1: The Impact of Clean Air Act Regulations and OTAG Controls ................ 26
Policy Analysis 2: Comparison of the CAC Scenarios to Scenarios for Achieving
Nitrate Loading Reductions ..................................................................... 34
Policy Analysis 3: National Policy and the Bay Community's Stake ...................................... 50
VII. Conclusions and Policy Implications .................................................................................... 58
A.
Regulatory Analysis (CAA and OTAG Scenarios) ...................................................... 59
B.
Comparison of OTAG Regulatory Scenario to Alternative Scenarios for
Meeting Nitrate Loading Targets ................................................................................ 60
C.
The Bay Jurisdictions' Stake in NOx Trading for Reducing Ambient Ozone ............... 62
D.
Limitations and Caveats ............................................................................................. 63
Appendix 1: The Convex Hull and Optimization ............................................................................ 65
Appendix 2: Utility Cost Functions ................................................................................................ 68
Appendix 3: Mobile Source Data .................................................................................................... 69
References ..................................................................................................................................... 73
List of Figures and Tables
Figure 1.
Figure 2.
Figure 3.
Figure 4.
Figure 5.
Figure 6.
Figure 7.
Table 1.
Table 2.
Table 3.
Table 4.
Table 5.
Table 6.
Table 7.
Table 8.
Table 9.
Table 10.
Table 11.
Watershed and Expanded Airshed Outlines .................................................................. 6
Chesapeake Bay Modeling Domain ............................................................................ 17
Average wind pattern for 5-day Northeast episode ...................................................... 19
Difference in NOx Emissions by State: LC compared to CAC Scenario ..................... 38
Cost Effectiveness of Nitrate Reductions by State ...................................................... 39
Difference in NOx Emissions by State ....................................................................... 41
Cost Effectiveness of Nitrate Reductions .................................................................... 42
Load to Emission Ratios ............................................................................................. 15
Population weighted NOx to ozone S-R matrices, by source type and episode ............ 22
Trading Ratios, by source type and episode ................................................................ 24
Annual Unit Damages for Health Effects from Ozone ................................................. 25
Monetary Benefits over the Study Domain per ton of NOx Emitted ............................ 25
No-Control Baseline NOx Emissions and Nitrate Loads. Year 2005 ........................... 27
Share of NOx Emissions and Nitrate Load Reductions by Source Category ................ 27
Annual Nitrate Load and NOx Emissions Reductions, Benefits and Costs ................... 28
Gross and Net Costs per pound of Nitrate Reduced ..................................................... 28
State Contributions of NOx Emissions and Nitrate Loading to the Bay from
Air Sources ................................................................................................................ 32
Costs of NOx Emissions and Nitrate Loading Reduction ............................................ 33
iv
Table 12.
Table 13.
Table 14.
Table 15.
Table 16.
Table 17.
Table 18.
Table 19.
Table 20.
Summary Results for OTAG and Least Cost Scenarios ............................................... 36
Most common technologies in mobile source regions ................................................. 44
The Most Common Technologies in CAA, OTAG, and Least Cost (OTAG)
Scenarios .................................................................................................................... 46
Effects of Different Ozone Episodes on Results from Net Least Cost
Scenario ..................................................................................................................... 47
Summary Results from OTAG and Least Cost Scenarios for the Bay States ................ 48
Consequences of Alternative NOx trading plans to Meet Aggregate NOx
Emissions Reduction Targets ...................................................................................... 52
Consequences of Alternative Trading Ratios for Meeting an Ozone Exposure
Reduction Target ........................................................................................................ 55
Loss in Benefits from a Mismatch between Trading Ratios and Episode
Experienced ............................................................................................................... 57
Single zone versus two-zone trading to meet ozone exposure targets ........................... 57
v
THE CHESAPEAKE BAY AND THE CONTROL OF NOX EMISSIONS:
A POLICY ANALYSIS
Alan Krupnick, Virginia McConnell, David Austin,
Matt Cannon, Terrell Stoessell, and Brian Morton *
I. INTRODUCTION
After years of worsening water quality in the Chesapeake Bay caused primarily by
eutrophication from excessive nutrient inflows, the Governors of Maryland, Virginia, and
Pennsylvania, and the Mayor of D.C. signed the Chesapeake Bay Agreement in 1987.1 This
agreement required 40 percent reductions in the controllable fraction of nutrients reaching
the Chesapeake Bay by the year 2000. Municipal treatment plants and agricultural runoff
were thought to be the primary sources of such nutrients. By 1992, bans on phosphorus in
detergents and improvements in municipal sewage treatment, as well as some voluntary
controls on farm runoff and land use restrictions helped bring about a 20 percent reduction
in phosphorus loadings, but only a 4 percent reduction in nitrates. The relatively recent
finding that emissions of nitrogen oxides (NOx) from the air are a major source of nutrient
enrichment in the Chesapeake Bay, comprising anywhere from 20 percent to 35 percent of
the Bay's controllable nitrate loads (Dennis, 1997), sparked major interest in pursuing NOx
emissions reductions for Bay improvement.
During the same period, NOx emissions in the eastern U.S. have become the focus of
efforts to meet ambient air quality goals. The Clean Air Act Amendments of 1990 (CAAA)
initiated several NOx emissions reduction policies to meet both the ambient ozone standard
(because NOx is an ozone precursor along with volatile organic compounds (VOCs)), and as
part of the program to reduce acid deposition of sulfur dioxide (SO2) and NOx emissions
from utilities. In recognition of the importance of long-range transport of ozone in the
eastern U.S. and the crucial role played by NOx in ozone formation, the Ozone Transport
Assessment Group (OTAG) and the subsequent "Proposed NOx Trading Guidelines" issued
by EPA supported the regional trading of NOx emissions to reduce ozone cost-effectively.
The recent EPA rules tightening the ozone standard and setting a new fine particle standard
* Alan Krupnick, Senior Fellow and Acting Division Director, Quality of the Environment Division, Resources
for the Future; Virginia McConnell, Senior Fellow, and David Austin, Fellow, Resources for the Future; Matt
Cannon, Research Associate and Terrell Stoessell, Research Assistant, Resources for the Future; Brian Morton,
EC/R Inc.
1 Chesapeake Executive Council, 1987.
1
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
will put further focus on NOx emissions reductions.2 The possibility that the Chesapeake
Bay may be the recipient of these "costless" nitrate loading reductions3 as a result of both the
1990 Clean Air Act and the new EPA ambient air quality standards suggests that these
benefits should be quantified.
This paper takes several perspectives on the issue of nitrate loadings to the Bay
coming from airborne sources. It examines the costs, emissions, source-specific and locationspecific allocations of NOx emissions reductions, and the ancillary ozone-related health
benefits associated with NOx emissions reductions under a range of policy scenarios. These
include command and control scenarios -- those applicable to states responding to the Clean
Air Act Amendments of 1990 and to recent suggestion for further NOx emissions reductions
made by OTAG. They also include scenarios designed either for meeting nitrate loading
reduction targets or ozone exposure reduction targets in a cost-effective manner. The former
target is examined as well in a model that permits ancillary ozone-related health benefits to be
factored into the abatement allocation decision.
Specifically, this paper addresses three different issues:
Policy Analysis 1. The Impact of Clean Air Act Regulations and OTAG Controls. We
examine the impact of NOx emissions reductions, which are anticipated to occur under
existing air regulations targeted toward meeting ambient ozone standards, but which will also
reduce nitrate loading to the Bay. Two separate "command and control" scenarios are
considered: a Clean Air Act (CAA) scenario and additional mandated controls anticipated as
part of what we term the OTAG scenario (one aspect of which is the reduction of NOx by
electric utilities to meet a 0.15lb NOx/mmBTU performance standard; see below). We
examine the impact of these regulations on nitrate loading and give estimates of both gross
costs and costs net of the ozone-related health benefits.
Policy Analysis 2: Comparison of the OTAG Command and Control Scenario to
Other Scenarios for Achieving Nitrate Loadings Reductions. This analysis takes the CAA
scenario as the starting point and compares the outcomes from the OTAG command and
control scenario to various alternatives:
2 The implications of NOx emissions reductions for fine particle concentrations will be examined in a
subsequent analysis. Though only 3-6% of the inventory of fine particles in the eastern U.S. is thought to be
nitrates (formed through the conversion of NOx to nitrates in the presence of ammonia in the air), with a far
higher fraction (30-40%) thought to be sulfates (from SO2 emissions), the health benefits of NOx control
realized through the nitrate channel are estimated by EPA to outweigh the health benefits realized through the
effect of NOx emissions reductions on ozone concentrations.
The most recent report from the Interagency Monitoring of Protected Visual Environments (IMPROVE)
suggests a different composition for the inventory. For the Washington DC area, fine particles of nitrates are
15.1% in the spring, 5.6% in the summer, 12.8% in the fall, and 21.5% in the winter. Sulfates make up 46.6% of
the concentrations in the spring, 60.4% in the summer, 42.7% in the fall, and 32% in the winter. While DC
pollution is not necessarily representative of the entire Chesapeake Bay airshed, these numbers could be used as
an upper limit of what to expect from fine particles in the immediate Chesapeake Bay region (Sisler, 1996).
3 Costless from the point of view of regulatory activity associated with the Bay.
2
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
a. To the extent that the Chesapeake Bay policy community has an interest in
focusing efforts for NOx emissions reductions on improvements in the Bay, it
would be of interest to know the least cost way of getting nitrate loadings
reductions and how this would compare to using the command and control (CAC)
approach. Alternatively, the Bay community might want to include the
maximization of ozone reduction benefits that stem from NOx emissions
reductions into its objectives, having this factor as well as nitrate loading goals
influence the allocation of NOx emissions reductions.
b. In another scenario, we examine the case in which no further NOx emissions
reductions would be forthcoming in the Chesapeake Bay airshed after those
specified in the CAA scenario and that, under these conditions, the Bay
jurisdictions decide to "go it alone." That is, the Bay community on its own might
focus on cleaning up the Bay, seeking NOx emissions reductions in the Bay
jurisdictions for reducing nitrate loading. Or the community might want to focus
on both nitrate loading and ozone benefit objectives. Given what they might
otherwise have to do in an OTAG scenario, how would such other, more local
policies compare?
Policy Analysis 3. National NOx Policy and the Bay Community's Stake. EPA's
efforts to design a NOx trading program for the eastern U.S. have implications for the
jurisdictions in the Chesapeake Bay Agreement. This analysis examines whether there are
some design choices that should be favored or opposed by the Bay jurisdictions, according to
the effect these choices have on nitrate loading, the health of the population living in the Bay
jurisdictions, and the abatement costs to NOx sources located in the Bay jurisdictions. This
analysis, by focusing on a 13-state domain including DC, also provides implications for the
overall design of a NOx trading system.
II. PREVIOUS RESEARCH
An initial attempt to quantify the reduction in nitrate loading that might be realized
from the Clean Air Act and some additional policy initiatives was made by Pechan et al.
(1996). Pechan began with a 2005 emissions inventory developed as input to the EPA's
Regional Oxidant Model (ROM) and this baseline was reduced according to assumptions
associated with the implementation of the technology-based elements of the 1990 CAAA and
some further reductions associated with new initiatives in discussion at that time.4 The cost
4 The OTC-Low Emission Vehicle petition and a possible requirement that electric utilities reduce NOx
emissions to the lesser of 0.15 lbs NOx/million Btu boiler heat input or 85% reduction and other large point
sources meet the same rate reduction or 70% reduction, whichever is lower. Scenario "C2" applied these
restrictions to the Bay jurisdictions alone. Scenario "E" applied them to the entire Bay airshed, defined in Dennis
(1996) as the geographic area whose NOx emissions account for 75% of the RADM model's predictions of Bay
watershed deposition.
3
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
functions relating NOx emissions reductions to costs for pollution controls were estimated for
electric utilities, other point sources, and mobile sources, taken from Pechan's Emission
Reduction and Cost Analysis Model for NOx (ERCAM-NOx). The effects on nitrate loading
were examined by pairing source-receptor coefficients that convert NOx emissions to nitrate
deposition by Bay sub-basin (taken from runs of the Regional Acid Deposition Model
(RADM)) to coefficients from the Chesapeake Bay Watershed Model, that in turn estimate
the share of deposition by sub-basin to reach the Bay as nitrate loading.
Pechan (1996) found that the Clean Air Act applied to NOx sources in the Bay airshed
states (Delaware, D.C., Indiana, Kentucky, Maryland, Michigan, New Jersey, New York,
North Carolina, Ohio, Pennsylvania, Tennessee, Virginia, and West Virginia) resulted in
reductions of 11.6 million pounds of nitrate loading at an average cost per pound reduced of
$123. A scenario that examined additional controls in the Ozone Transport Region brought
about a further 5.6 million pounds of reductions, raising the average costs to $147 per pound.
Utility reductions were by far the cheapest, at $95/pound, versus $329 per pound from Low
Emission Vehicles (LEVs) and $466 per pound from non-utility point sources.
Because NOx emissions reductions within the Bay jurisdictions are more productive
for reducing nitrate loading to the Bay than are reductions from further away, the cost per
pound for the Clean Air Act scenario applied only to the Bay jurisdictions is $75, with utility
reductions in NOx the most cost-effective. A state like Kentucky was found to reduce a
pound of nitrates, other things equal, at almost six times the cost of reducing a pound in the
state of Maryland.5 Overall, however, reductions in NOx emissions were found to be costcompetitive for reducing nitrate loading when compared with at least some forms of waterbased loading reduction costs, even for air sources within the entire airshed. For instance,
urban non-point source reductions were estimated to cost $143 per pound nitrates, although
for some agricultural practices (such as low-till, which utilizes less fertilizers than standard
tillage) costs drop to about $7 per pound (compared to the lowest estimate for any airborne
source-state combination of $39 for Maryland utilities).
There are a number of ways the Pechan analysis can be improved, updated and
extended. First, the Pechan analysis attributes all costs of air NOx emissions reductions to
nitrate loading declines in the Chesapeake Bay. However, most of these reductions have been
done or will be undertaken to improve air pollution and the associated health benefits, so
attributing all of the costs to nitrate reductions in the Bay overstates the costs. At a minimum,
if the costs of reducing nitrate loading to the Bay are attributed to the Bay cleanup, at the least
the ancillary benefits from ambient ozone or other air quality improvements should first be
subtracted from these costs. We show in the theoretical section below how and why the air
quality improvements should be netted out of the costs of nutrient reduction. This netting out
5 For utilities only, Maryland utilities meet the 0.15 lb/mmBtu restriction at a cost of $1,200 per ton NOx while
Kentucky utilities meet it at $1,100. In terms of nitrate loadings, Maryland utilities face a cost of $39 per pound
while Kentucky's utilities register at $254 per pound.
4
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
could not only significantly lower the average cost of nitrate loadings reductions from the air
but will also be likely to alter the geographic and source-specific cost per pound comparisons.
Second, the analysis only considers existing command and control measures for
reducing NOx. It does not examine alternative economic incentive control options that might
achieve the same goals, but at lower overall costs. This may be important in the future, as
more stringent policies continue to push the additional costs of control higher. Also, in light
of the support for NOx trading from OTAG participants and the EPA, it would be useful to
also consider the level and distribution of costs per pound associated with a tradable NOx
permit scheme. This could be compared to a tradable ambient nitrate loading permit scheme.
The latter is equivalent under certain conditions to a least cost allocation of NOx emissions
reductions to meet any given nitrate loading reduction goal.
Finally, the Pechan analysis used emissions and cost estimates which Pechan has since
updated, as part of its work in developing the OTAG inventory. Thus, the purpose of our
analysis is to extend the initial Pechan analysis by examining the impacts and costs of both
command and control and economic incentive (or lower cost) scenarios for reducing nitrate
loading and NOx emissions, capturing the joint health benefits associated with ancillary
reductions in ozone, and to do so using more up-to-date emissions and cost data.
III. DOMAINS AND SCENARIO DEFINITIONS
A. Domains
The NOx emissions sources included in the analysis are located in the Chesapeake
Bay airshed, as defined in Dennis (1996) using the Regional Atmospheric Deposition Model
(RADM). This domain is shown in Figure 1 (along with the Chesapeake Bay watershed).
Nitrogen oxides are emitted from sources in the airshed and transformed to nitrates. A
portion of the nitrates reaches the watershed, is deposited to the land surface, of which a small
fraction is released to the streams and rivers and the resulting nitrates are transported to the
Chesapeake Bay. Some deposition is also directly to the Bay itself. Dennis shows that nearly
every area in the U.S. east of the Mississippi River contributes something to the airborne
nitrate deposition to the Bay watershed. But many regions have contributions too small to
warrant inclusion in a major reduction effort. To focus on those regions that have the greatest
impact, Dennis identified the areas that account for 75 percent of the watershed deposition.
Based on this analysis, we included all or part of 13 states and the District of Columbia in our
Chesapeake Bay source domain.
Because our analysis includes the effects of NOx emissions reductions on ambient
ozone and on the health of people living in areas where ozone concentrations change, the
domain for estimating these effects includes the source domain and any additional areas in the
U.S. likely to experience significant ozone changes as a result of NOx emissions reductions in
the source domain. These additional areas include Connecticut, Long Island, Massachusetts,
and Rhode Island.
5
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
Figure 1
6
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
B. Scenarios
Each scenarios is defined with emissions control relative to a baseline. We first
discuss the possible baselines.
Baselines
The baseline for the CAA scenario is 1990 emissions projected to 2005 using Bureau
of Economic Analysis (BEA) and other projections of activity levels. This baseline does not
account for emissions reductions mandated in the CAAA of 1990. The baseline for all other
analyses is the NOx emissions, abatement measures, and activity levels projected by Pechan
& Associates for 2005 under the Clean Air Act Amendments of 1990. Pechan (1996) defines
the CAA scenario by applying growth and control factors to the Interim Inventory (EPA,
1993). These control measures include: Reasonably Available Control Technology (RACT)level NOx controls on major point sources in ozone nonattainment areas, Title IV (Acid Rain)
NOx emissions reductions on steam-electric utilities, enhanced inspection and maintenance
(I&M) programs in the more polluted ozone nonattainment areas (under Title b), federal Tier I
vehicle emissions standards (under Title II)), and stringent controls on projected new major
point sources in ozone nonattainment areas (assumed to be selective catalytic reduction)
required under EPA's New Source Review requirements.
Scenarios
There are three sets of scenarios that follow the major policy analyses presented in the
paper. The first set are the command and control (CAC) scenarios; the second set are
alternative scenarios that have the same goal of Bay nitrate loading reduction as the CAC
scenarios, which we term the least-cost and emissions trading scenarios. The final scenarios
deal with the design of possible NOx emission trading policies for meeting ozone reduction
goals in the eastern U.S. We describe each of these briefly below.
1. Command and control scenarios
There are two command and control scenarios. The first is basically the Pechandefined Clean Air Act scenario, referred to here as CAA. The second (OTAG) includes
higher levels of control identified as part of the OTAG process. These were 0.15 lbs
NOx/mmBTU of boiler heat input, or 85 percent NOx removal for utilities and 70 percent
NOx removal for non-utility point sources, whichever is less. EPA subsequently used
approximately the same standards to undergird its State Implementation Plan (SIP) call to the
states (FR, 1997).
Mobile sources were virtually ignored in the OTAG process. However, the original
Chesapeake Bay Cost Allocation Study (Pechan, 1996) examined the effects of implementing
a Low Emissions Vehicle Program in the Ozone Transport Commission Region (the
Northeastern States). We require LEVs in all areas in our OTAG scenario.
7
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
2. Alternative scenarios for achieving Bay nitrate loading reductions
There are a number of scenarios in this part of the analysis, all of which take as their
goal nitrate loading reductions that are equivalent to those resulting from the OTAG scenario.
2a. Minimum costs for the entire Airshed
•
We examine first the allocation of NOx emissions reductions among sources and
regions that would provide the lowest cost way of meeting the nitrate loading
reduction target, attributing all of the costs to nitrate loading reductions.
•
An extension of that scenario is to include the health benefits realized through the
impact of NOx emissions reductions on ozone concentrations as part of the cost
minimization algorithm. We examine both "mid" and "high" estimates of those
ozone benefits (see below).
•
Another scenario examines the allocation of reductions if the costs of NOx
emissions reductions were minimized, while still achieving the same nitrate
loading reduction goal. In this case, the allocation of NOx emissions proceeds by
equalizing marginal costs per unit of NOx emissions reductions rather than the
marginal costs per unit of nitrate loading reductions. One can think of this
scenario as a trading system, like that of the S02 allowance market, where
emissions trade on a one-to-one basis. Such a system will be less efficient than an
ambient system where the effects of a ton of NOx on loadings are different
depending on the location and type of source. However, there is experience with
such a system and at least the perception that an emissions trading system has
lower education and transactions costs than an ambient permit system.
2b. Bay states undertake low cost policies independently
•
Finally, we include scenarios that focus on nitrate loading reductions only from the
Bay jurisdictions,
The least-cost scenarios can also be thought of as similar to ambient trading programs.
For instance, one can envision a perfectly operating allowance trading market where the
currency of the trade is nitrate loadings to the Bay and a cap is set on nitrate loadings equal to
that predicted by the OTAG scenario. This type of scenario has been shown to yield
equivalent results to the least cost allocation of emission reductions to an ambient goal, in this
case, in terms of nitrate loadings (Montgomery, 1972).
This system implies that sources whose NOx emissions have differential effects on
nitrate loadings will trade their NOx emissions at ratios different than one. For instance,
assume an electric utility in Ohio buys nitrate loading permits from an electric utility in
Maryland, but the Ohio utility's NOx emissions have only half the effect on Bay loadings as
8
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
that of the Maryland utility. Then, the Ohio utility could increase its NOx emissions by twice
the reduction in NOx emissions from the Maryland utility, without changing loadings to the
Bay. Thus, the trading ratio in terms of NOx emissions between these sources would be set at
2:1.
All of these scenarios are to some extent hypothetical, since air pollution policies are
unlikely to focus primarily on Bay clean-up goals. They are informative, however, in that
they help identify the lowest cost sources of additional reductions.
Note also that, in both the CAC and the alternative scenarios for meeting nitrate
loading goals, NOx emissions reductions yield benefits from reduced ozone exposures that
may be considered ancillary to the goal of improving the Bay. All analyses that involve
meeting nitrate targets include estimates of these ancillary benefits. Because of the wide
confidence intervals associated with health benefits estimates, we also perform some
sensitivity analyses with 5 percent and 95 percent confidence interval estimates of benefits per
ppb ozone-person-day (see below for a full discussion about how these benefits are
estimated). In addition, the ozone benefits from NOx emissions reductions can be included
directly in the NOx emissions allocation decision, with the effect that the allocation seeks to
meet the nitrate loading reduction goal while minimizing the net costs of NOx control, which
is the gross costs minus the ancillary ozone benefits (see the next section for the formal
model).
3. Alternative scenarios for reducing ozone exposures
In this part of the analysis, we are interested in whether some options for NOx trading
for ozone improvements are more or less favorable to the Bay jurisdictions than other options.
Along the way, we examine the implications for abatement costs across the airshed states.
First, we consider the OTAG regulatory allocation of NOx emissions reductions and
compare it to various systems of emissions trading, including EPA's plan that limits trading
on a one-to-one emissions basis and only between utilities. We then examine variations on
EPA's emissions trading plan, adding additional source types to trading and considering plans
that would trade emissions at ratios guided by the effect of a source at a given location on
ozone exposures to the population in a receptor region. The Bay jurisdictions' interests are
described in terms of nitrate loading reductions, ozone exposure benefits, and abatement costs
to Bay jurisdiction sources. We then address questions about the sensitivity of the results to
alternative meteorological regimes. Finally, we examine a zonal trading plan where the Bay
jurisdictions define their own trading zone in which more attention can be given to nitrate
loading reductions.
In contrast to the scenarios for nitrate improvements, these scenarios have direct policyrelevance to the shape of a future NOx trading plan. No government officials (but some
academics) are discussing the possibility of trading against an ozone exposure reduction goal.
All analyses provide estimates of the annual costs and benefits for each scenario
compared to some baseline for the year 2005, expressed in 1990 dollars.
9
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
IV. THE CONCEPTUAL MODEL
This section presents the conceptual model for allocating NOx emissions reductions
cost-effectively to meet a nitrate loading reduction goal and, separately, an ozone exposure
reduction goal. In addition, it specifies how the former goal can be met while simultaneously
accounting in the optimization for the ancillary ozone-related health benefits associated with
NOx emissions reductions.
Adopting notation used in Tietenberg (1985), in the least-cost model we seek to
minimize total control costs over J sources of nitrogen oxides,6 given that we must reduce
total Bay nitrate loading to some pre-established limits or targets Wi at each of I measurement
locations. If loadings are uniformly mixed, as we assume for the Chesapeake Bay, there will
be only a single location and one limit W , but for generality we assume multiple locations.
Letting c j (rj ) represent the cost of reducing NOx emissions at source j by the amount
r j -- from some baseline emissions level e j -- and letting the function L ji (⋅) map NOx
emissions at source j into Bay loadings at site i, the objective is simply to minimize the
aggregate cost of those reductions across all sources, subject to achieving the water quality
targets. The objective is:
J
min C ( R) = min ∑ c j (r j ),
rj
rj
(1)
j =1
subject to
− Wi + ∑ L ji (e j − r j ) ≤ 0
J
(2)
j =1
That is, the regulator seeks to find the least expensive way of assuring that loadings
L ji of the emissions remaining after controls are installed, (e j − r j ), are no greater than the
imposed water quality target Wi . This model is easily adaptable to the case of meeting some
target X for aggregate ozone exposures per "ozone season."7 In that case, X would
substitute for Wi in (2), and a set of source-receptor coefficients djk mapping emissions at
source j to ozone exposures at air receptor location k would replace the loadings matrix
elements L ji .8
6 It is an open question how those J sources are selected. An extension of our work would be to derive the
optimal extent of a regulatory region. This would involve trading off between regulator power, which
presumably is greater the fewer sources that must be controlled, and costs of environmental control--as
extending the regulatory domain could lower costs by creating increased gains from trading over a larger market,
as well as possibly bringing large emissions sources under the control of the regulator.
7 This is the sum of ozone parts-per-billion person-days over the entire airshed and ozone season.
8 Because the ozone target is an aggregate one, the source-receptor coefficients djk would be summed both over
sources j and receptors k, and there would be a single standard.
10
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
In either case, total costs, summed over all emissions sources, are given by C(R). To
achieve an optimum given that reductions rj must be non-negative -- as must be the
LaGrange multiplier λi on the environmental target in (2), -- the Kuhn-Tucker conditions
(see, e.g., Varian, 1992) must be satisfied. When the Bay loadings serve as the constraint, the
key first-order condition for cost minimization, which we explain below, is:
∂c j
∂r j
I
− ∑ λ i L ji ≥ 0
(3)
i =1
Cost minimization will not be achieved unless a "complementary slackness" condition
is also satisfied, guaranteeing that for every source that optimally reduces its NOx emissions
(i.e., for which rj>0), condition (3) is met with equality. In other words, at the optimum point,
all sources are controlled to where their marginal control costs equal the sum of the shadow
prices (the marginal cost of the last unit of pollution reduction) λi of the environmental
targets, weighted by the loadings Lji of their NOx emissions at each Bay receptor. For the
single receptor (or uniform mixing) case, (3) implies that sources are controlled to where the
ratios of their marginal control costs, MCj, to their Bay loading factors Lj are all equal:
MC j
Lj
=λ =
MC j '
Lj'
for any sources j and j'. Sources with marginal costs too high to achieve this would not reduce
their emissions.9
This outcome can be achieved by regulatory fiat, but it has been shown elsewhere that
a market-based, tradable permit approach, where permits for emissions, producing total
loadings of Wi at each receptor i, can achieve the same outcome. (See Montgomery (1972)
for initial development of a trading model in this context, and Krupnick et al. (1983) for
refinements.) The same is true for the case of controlling emissions to meet an ozone
exposure reduction target.10 In a competitive trading market, equilibrium permit prices of
Pi=λi would arise at each (Bay or air) receptor site and, subject to the qualifications in
Krupnick et al. (1983), would achieve the cost-minimizing outcome without regulatory
intervention (beyond issuing the correct number of permits and setting the correct trading
9 For the case of an aggregate ozone exposure reduction goal, there would be only a single shadow price .
K
∑D
j
Each source's loading coefficient Lj in this expression would be replaced by the source's contribution k =1
to
total ozone exposures.
10 In our least-cost simulations, we will take from the command-and-control case the ozone exposure reductions
implied, and treat these reductions as the goal.
11
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
ratios). That is, separate markets would exist for permits specific to each receptor.11 Or said
another way, trading ratios would need to be set for trades between any two regions. The
initial allocation of the permits need have no effect on the outcome. Under this scheme, lowcost sources would reduce emissions beyond what they are permitted to emit, and would sell
their excess permits at price Pi to high-cost sources.
The control of NOx sources to achieve water quality goals will inevitably also create
benefits from improved air quality. These "ancillary" air benefits must be accounted for if
emissions controls are to achieve the water quality target in a socially optimal manner. Here
the problem is one of social welfare maximization, rather than cost minimization, given an
imposed water quality target. That is, we seek to maximize ancillary air benefits Ba net of the
costs of achieving the target:
 J K
 J

⋅
max Ba  ∑∑ d jk r j  − ∑ c j (r )
rj
 j =1 k =1
 j =1
(4)
subject again to constraint (2). Here djk is a source-receptor matrix mapping emissions at
source j to concentrations at (air) receptor k; and cj and rj are as before. (Note the substitution
of "concentrations" for "loadings" when the constraint switches from water quality to air
quality.) The first-order condition for maximization of (4) which is analogous to the first
order condition (3) is:
∂Ba
∂r j
K
∂c j
k =1
∂r j
∑ d jk −
I
+ ∑ λi L ji ≤ 0
(5)
i =1
As before, complementary slackness implies that the optimal point of control occurs when
marginal control costs net of marginal ancillary benefits are equal to the weighted shadow
prices of the constraints. This can be seen by rearranging terms in (5):
I
MC j − MBa j = ∑ λi L ji .
i =1
( MBa j is the marginal ancillary benefits, (∂B j ∂r j )∗ ∑ d jk , from controlling source j.)
K
k =1
Where uniform mixing is assumed, this condition simplifies to
MC j − MBa j
Lj
=λ=
MC j ' − MBa j '
(6)
L j'
11 If different source types affect pollution transport (or exposures) differently, e.g., utilities with tall stacks
create more ozone downwind than mobile sources, the source-receptor matrices would differ by source type as
well as location. This implies that trading ratios in a marketable permit system, to maximize the gains from
trade, should also differ by source types involved in the trade, whether or not the trade crosses regional
boundaries.
12
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
for any two sources j and j'.
Three important results are implied by this equation. First, rather than controlling all
sources to where their marginal control costs equal Ljλ, the shadow price of the constraint
weighted by their individual loading factors, here sources which produce air benefits are
controlled more than they otherwise would be, as an increasing function of those air benefits.
Second, accounting for the air benefits implies a lower shadow price λ than before, one that
reflects the true social cost of meeting the water-quality target. Finally, and perhaps most
important, because the air benefits are external to the Bay clean-up, a trading market for Bay
nitrates would not achieve the optimal outcome unless a mechanism were implemented to
internalize each source's ancillary benefits in the abatement decision. The straightforward
system for the gross cost-minimization model would not work here.
V. DATA
A. Emissions and Costs
The basic database for emissions and costs in the analysis which follows is contained
and generated in a model developed by E. H. Pechan and Associates for EPA, called the
Emission Reduction and Cost Analysis Model for Oxides of Nitrogen (ERCAM-NOx). This
model is supplemented by VOC emissions data for mobile sources (needed for projecting
ozone effects) and a few cost revisions to the E.H. Pechan cost assumptions.
ERCAM-NOx (1996) contains 1990 baseline NOx emissions for utilities, other point
sources, mobile sources, and other area and nonroad sources and projects these emissions
forward to 2005 using state-level, 2-digit Bureau of Economic Analysis earnings indicators
and population projections. Cost and associated emission reduction estimates were developed
for a variety of "engineering" abatement options for each source-type.12 Each point source is
assigned a unique vector of abatement costs based on the size of that source's boiler and
associated economies of scale. Finally, specific abatement options by source are identified
that would be in place in 2005 in response to requirements set by the 1990 Clean Air Act and
its implementation as specified in a state's Implementation Plan.
Interested readers should refer to Pechan (1996) for details. Note that mobile source
emissions are estimated at the county level by using county-specific data as input to the
MOBILE5 Model, EPA's current generation vehicle emissions inventory model. The county
level data include ambient temperature, the type of Inspection and Maintenance (I&M)
program in place or forecast, and the type of reformulated gasoline to be used.
Developing reasonable estimates of the costs of mobile source abatement technologies
for 2005 is difficult. Technological and political changes make the evolution of particular
controls, such as LEVs and even enhanced I&M programs, challenging to forecast. The current
Pechan database relies on state-by-state forecasts for the Clean Air Act scenario, but some of
12 Facilities covering about 23% of the emissions from non-utility point sources have no abatement options in
the Pechan database.
13
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
those forecasts are already proving to be inaccurate. In addition, the emission reduction
forecasts in the Pechan database for certain mobile source strategies are assumed by EPA's
MOBILE inventory model to be quite large. For example, the combination of LEV vehicles
and enhanced I/M reduces NOx emissions by more than 25 times the reduction from a basic I/M
program. These large forecast emissions reductions for the enhanced I/M alternatives come
from the MOBILE Model which we think may be overstating the reduction potential of
Enhanced I/M (Harrington, McConnell and Cannon, 1998). However, in this analysis, we use
the Pechan data for mobile sources including the most recent cost estimates used by E.H.
Pechan. In future work, we will use our own estimates of costs and emission reduction under
different mobile source strategies and compare these to the Pechan estimates. Appendix 3
describes the methods for predicting cost in more detail. Finally, because mobile sources emit
both NOx and VOCs, both of which contribute to the formation of ozone, we use the average of
the NOx and VOC reductions as the "NOx emissions reduction" for calculating ozone benefits.
A particularly challenging issue in determining the cost of control from both point and
mobile sources is how to count the costs of moving from one technology to another. If, for
example, technology A is in place in the CAA scenario, but technology B would be used in
the OTAG scenario, what are the costs of the B technology, given that A is already in place?
There are a number of situations in which the costs are easy to determine -- if the B treatment
can clearly be added on to the technology in place, then the costs of moving to B are just the
costs of B. An example for utility sources is controls for wall-fired coal boilers. Low-NOx
burners can be installed alone or in combination with overfire air (the "B" technology) to
reduce NOx. The difference in cost between the two is the cost of the overfire air technology.
For mobile sources, if reformulated gas is added to I&M which is already in place, the added
costs are just the additional cost of the reformulated gas.
There are, however, many situations where the new controls require a switch from one
technology to another, rather than an add-on. In the case of mobile sources, switching from
one I&M type to another, or in the case of utilities, wall-fired coal boilers can change from
low-NOx burners to selective non-catalytic reduction. In these situations we assume that the
capital cost of the dropped technology is not recoverable for point sources and that 25 percent
of the capital cost is recoverable from I&M programs. We also assume that the O&M costs of
the dropped technology are fully recoverable.
B. NOx to Nitrate Source-Receptor Coefficients
Outputs from runs of EPA's Regional Acidic Deposition Model (RADM) and the
Chesapeake Bay Program's Watershed Model (CBWM)13 are linked in Pechan (1996) to
13 The CBWM is used to relate a kg of nitrate deposition by basin to nitrate loadings actually reaching the
Chesapeake Bay. The CBWM (Phase III) (Linker, et al., 1993) simulates the land and water movement of
nutrients from the Bay sub-basins to the Bay for a wide variety of nutrient sources, including airborne sources.
The modeling domain covers the 64,000 square miles of the Chesapeake Bay's drainage system. This model,
calibrated to weather conditions and water flows over the 1984-87 period, estimates total annual loads to the Bay
of 324 million pounds of nitrates, of which 27%, or 105 million pounds, come from airborne sources.
14
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
convert NOx emissions at a source location (either substate, state or region) and for a source
type to deposition in one of fifteen Chesapeake Bay Basins (in kg of wet plus dry nitrates per
hectare per year), and deposition in these basins to nitrate loadings to the Chesapeake Bay (in
pounds per year). Table 1 provides load to emission ratios by state. As a point of departure,
our estimate of loadings from the point and mobile sources in our domain (which omits area
and non-road sources and is for much lower NOx emissions than implied by the CBWM
scenarios) is about 66 million pounds for the base year.
Table 1. Load to Emission Ratios (lbs Nitrates / tons NOx)
Source
State/City
Atlanta
Delaware
Detroit
DC
Indiana
Kentucky
Maryland
New Jersey
New York
North Carolina
Ohio
Pennsylvania
East Pennsylvania
West Pennsylvania
Tennessee
West Virginia
Virginia
All
Utility
Mobile
Major Points
32.61
N
34.56
U
30.31
M
23.22
N
M
N
U,N,M
U,N,M
U,N,M
U,N,M
N
U,N,M
U,N,M
U,N,M
U
U,N,M
U,N,M
U
20.77
32.61
4.90
17.72
7.57
8.06
19.10
21.34
17.72
5.59
18.93
4.42
4.42
34.56
30.31
M
19.84
U
19.71
M
23.22
N
13.90
18.93
U,N,M
U
20.77
M
2.40
U,N,M
6.31
U,N,M
Codes:
U: This ratio was used for utility sources.
N: This ratio was used for non-utility point sources.
M: This ratio was used for mobile sources.
C. NOx to Ozone Source-Receptor Coefficients
One of the major areas of potential air quality benefits associated with reductions in
NOx emissions is health effects associated with reduced ozone exposure. To model these
health effects we linked NOx emission changes at point sources and NOx plus VOC emission
changes for mobile sources to ozone changes at receptor areas. These linkages were made by
running an ozone simulation model, the Urban Airshed Model (UAM-V), in what can be
called an "attribution mode" (Guthrie and Krupnick, 1998) for three "typical" ozone episodes.
The attribution mode involves developing a source-receptor matrix linking a source
type/region's NOx emissions to a change in ozone in each receptor region for a given set of
15
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
meteorological conditions. This is obtained by running the model for baseline emissions and
then perturbing each source region's NOx emissions in turn and attributing the change in
ozone over the grid to the source type/region and summing over the grids (in our case using
population weights) contained within each receptor region.
This approach may be starkly contrasted to the standard approach for using air quality
models: the scenario mode. In this mode, a baseline set of NOx emissions is fed into this
model which produces ozone concentrations. Then another set of NOx emissions is used, this
time related to a particular control scenario applied over the study domain. The resulting
changes in ozone are attributed to the NOx emissions reductions over the domain. However,
no attributions are made to specific source regions or sources. In addition, these analyses
usually are developed to examine nonattainment issues. Hence, the meteorological inputs to
the model are taken from multi-day episodes of extremely high ozone concentrations.
This scenario mode has two major deficiencies for estimating ancillary benefits in a
least-cost modeling framework such as ours. First, the restriction of episodes to extreme
violations of ambient standards is at odds with the needs of a benefits analysis that attempts to
measure annual or seasonal impacts. The former is performed to help judge when abatement
plans are sufficient to reach attainment. A benefits analysis is conducted to total the benefits
of season-long ozone reductions. Because it is generally agreed that there are no thresholds in
the effects of ozone on health, and that the concentration-response functions are linear, ozone
reductions from any baseline concentration generate benefits. Focusing only on extreme
events, which, by definition are rare, would vastly underestimate the ancillary benefits of
summer long ozone reductions. Rather, "typical" ozone episodes should be modeled to
capture health benefits over the ozone season.
Second, the scenario mode is completely impractical for finding a least-cost allocation
of NOx abatement. In this mode, hundreds, if not thousands, of trial and error runs would be
needed to identify a combination of abatement by source and region that attains the goal at a
minimum cost. On the other hand, if source-receptor coefficients linking NOx emissions at a
source-type and region to ozone concentrations at receptors can be specified, an optimization
model can be utilized, and the optimization takes a matter of seconds. Of course, sourcereceptor coefficients assume linearity or that the impacts of emissions changes are the same
regardless of the level of emissions. This assumption needs to be tested and we are in the
process of testing it now.
To use the UAM-V in "attribution mode," we defined six source regions (labeled 1
through 6 on Figure 2) and seven receptor regions (the same 6 source regions plus region 7:
New York City, Long Island, Connecticut, Rhode Island and Massachusetts). These regions
are small enough to experience no gross differences in winds within the region, are based on
physical distinctions affecting meteorology (separated by the Appalachian Mountains, for
example), and follow state lines as much as possible. We also needed a small number of such
regions to minimize the number of UAM-V runs required to develop the matrices. Finally,
receptor region 7 was added, even though this area is not a source of Chesapeake Bay nitrates,
to capture ozone-related effects of NOx emissions reductions within the source regions,
16
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
Figure 2
17
RFF 98-46
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
particularly New York and the Midwest. The resulting set of regions is by no means the only
possible choice, and the degree of distinction between regions will obviously vary with
weather conditions.
The overall approach to developing a particular "row" of a source-receptor matrix is to
perturb emissions from the sources in one source region (with inventories taken from ROM
inventories supplied by E. H. Pechan and prepared by SAI) and calculate the changes in daily
one hour peak14 ozone concentrations at each grid cell in the receptor regions and then
compute either an unweighted or population-weighted average ozone change over the grid
cells in a given receptor area, and then do the same thing for each source region in turn. This
approach (which is detailed, along with all the results) is described in Guthrie, Mansell, and
Gao (1998) and is a simplification of that used by Rao, Mount, and Dorris (1997).
We distinguish between S-Rs for NOx emitted at high altitudes, which would be
applicable to major non-utility sources and utilities, and S-Rs for mobile sources, which
perturb low-level NOx and anthropogenic VOC emissions by an equal percentage change.
NOx emissions from all point sources in a region were reduced by 70 percent of the 2005
baseline in the UAM-V runs for generating the point source S-R matrices. This figure is
within the range of expected NOx emissions reductions from point sources modeled within
the OTAG process and underlies EPA's recent proposed SIP call. For mobile sources, we
reduced NOx and anthropogenic VOC emissions by 20 percent from the 2005 baseline. This
is roughly in the range of expected reductions from mobile source controls discussed by EPA
and in the OTAG process.15
To derive the ozone source-receptor matrices, we decided to run the UAM-V for fiveday episodes and average the results for the last two days of each episode for each region.
The first three days are ignored to give the UAM-V a chance to work through the initial
conditions. Data availability dictated choosing episodes from 1990.16 We observed that there
are three roughly defined ozone patterns which are generally of interest. They are
characterized by broad but distinct areas of elevated ozone concentrations centered
approximately over the Midwest, the Northeast corridor and the Southeast. Figure 3 is the
average wind pattern for the 5-day Northeast episode we used. The episodes reflect in part
the physical distributions of precursor emissions and characteristic weather patterns, and in
14 Most of the concentration-health response functions use either this measure or an average over a longer
period.
15 Originally, a significant analysis of the non-linearities associated with the S-R coefficients was expected.
Thcse nonlinearities can arise from sensitivity to the ozone baseline or to the magnitude of the NOx (and VOC)
changes. Funding limitations precluded these tests. Note, however, that the statistical analysis by Rao, Mount,
and Dorris (1997) provides evidence that at a relatively high level of spatial aggregation, a linear model fits the
micro simulation results as well as any other.
16 The needed information consists of the output fields from a mesoscale meteorological model constrained by
observations. Most of the datasets which have been prepared for past ozone simulations fall into the worst-case
category, and/or do not cover the domain of interest, which is quite large by ozone modeling standards. The
only comprehensive dataset available which appears to be suitably prepared is the Interagency Working Group
on Air Quality Modeling (IWAQM) dataset prepared for the year 1990. Thc entire year was processed.
18
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
Figure 3
19
RFF 98-46
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
part the areas where exposure to elevated ozone concentrations has been identified as an air
quality issue. These patterns correspond (under extreme conditions) to the OTAG modeling
episodes, and more generally to the patterns identified in the recent Northeast States for
Coordinated Air Use Management (NESCAUM) report (Miller et al., 1997). This
correspondence is important because the NESCAUM report identified windfields for the
eastern U.S. that correspond to the top 20 percent of daily maximum ozone readings within
each of the three regions, in turn.17,18
A few caveats about these choices are in order. We recognize that 1990 was a "good"
ozone year, and that, therefore, our resulting source-receptor matrix may underestimate
typical ozone concentrations attributable to sources. Further, the large scale features of the
ozone patterns are likely to be well represented, but small scale features should be interpreted
with caution. In addition the absolute maximum ozone concentration recorded may not
capture the actual maximum for a particular day, since the observations are not evenly
distributed and may have missed the actual peak.
All told, we developed unweighted and population-weighted S-R coefficient matrices
for six source and seven receptor regions, for each of three ozone episodes, for both point
sources and mobile sources, summing to 12 matrices developed from 37 runs of UAM-V.
Table 2 presents the population weighted NOx to ozone S-R matrices along with "trading
ratios" (see below). Population data used to weight the ozone changes for areas within each
receptor region are taken from BEA 2005 projections by state.
To summarize the results, we consider first the average episodes. These are computed
by calculating a simple average of the population weighted two-day average source-receptor
coefficients for each of the three episodes, by source type and source-receptor combination.
We find:
(i) For both point and mobile sources: the effect of a region's NOx emissions on the
ozone exposure of its own population (the on-diagonal coefficients) is generally
larger than the transported effects, except New York NOx emissions affect ozone
in New England more than ozone in New York; indeed, transport effects are
17 The 20th percentile ozone concentrations are quite low: 70 ppb, 72, ppb, and 67 ppb for the NE, MW, and SE,
respectively. These concentrations are low because of the inclusion of rural monitors. Because of this, we feel
that urban concentrations, which for a health benefits analysis like ours are the most important target, are
representing most of the 20 percentile readings. Thus, by choosing intervals that generally mimic these 20th
percentile readings, we believe that we will reasonably capture typical urban summer day windfields when ozone
is high enough to be problematic.
18 The criteria for selection of candidate intervals were: 1) a relatively localized ozone peak (i.e., distinguishable
region of elevated ozone concentrations) in one of the regions of interest (northeast, midwest and southeast), 2) a
pattern duration of three days minimum, with five days preferable, 3) ozone concentrations above 80 ppb over a
broad region, with a preference for higher peak concentrations, and 4) preference for more strongly-defined
regional-scale wind flows, similar to those identified in the NESCAUM report. Six initial candidate intervals
were identified which were then narrowed to three. The selection criteria are necessarily somewhat subjective
and the selected set of intervals is certainly not unique.
20
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
relatively small in most cases, particularly from emissions in Tennessee and the South.
(ii) In particular, midwestern utility NOx emissions change ozone locally about three
times more per ton of NOx than in either New York or New England and far less
than the effect New York emissions have on New England ozone. Midwestern
utilities have a greater effect on Maryland and Virginia ozone than on that of New
England and New York.
(iii) There is evidence of a NOx "scavenging" effect. NOx scavenging is a physical
process by which reduced (increased) NOx emissions can increase (reduce) the
production of ambient ozone. Often, it occurs close to tall stacks emitting NOx
but the UAM-V model shows it occurring over large areas. In particular, we find
that reduced NOx emissions actually increase ozone on net for mobile source
NOx emissions from E. Pa.–N.J. on itself and on New England's ozone. The
"New England" effect counts ozone changes above Long Island and New York
City . This effect appears for each episode type.
Looking across episodes:
(i) The size of the coefficients is somewhat affected by episode type, although the
general points made above with respect to the average episode apply to the three
specific episodes.
(ii) One difference of interest is that for the classic northeast episode (i.e., a Bermuda
High), transport of emissions in the South and Midwest are more important. For
instance, Midwestern point source emissions affect New England, New York
City, and E. PA-NJ as much or more than these emissions affect the Midwest.
One relevant way to summarize these results is in terms of what are termed "trading
ratios." A trading ratio measures the total effect of a ton of NOx emitted in one region
divided by the total effect of a ton of NOx emitted in another region. Effects are measured by
ozone exposures (in ppb-person/day) and are computed by multiplying the ozone coefficients
in one column by the respective receptor region populations and summing the products. If
one were to develop a NOx trading program based on ozone exposure effects, these ratios
would define the terms of trade. Table 3 provides the trading ratios, for point and mobile
sources separately, by episode. The ratios are computed with the MD-VA region point
sources as the numeraire. This means, for instance, that reducing a ton of NOx from point
sources in New York results in 25 percent more benefits over the domain than doing the same
thing in Maryland-Virginia and that a ton of NOx reduced by mobile sources in New York
results in 22 percent more benefits than a ton of NOx reduced from point sources in
Maryland-Virginia.
21
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
Table 2. Population weighted NOx to ozone S-R matrices, by source type and episode.
Episode Type
Receptor
Region
Average
MD-VA
E.PA-NJ
MD-VA
E.PA-NJ
NY
OH-W.PA-WV
IND-KY
TN-South
NYC-LI-CT-RIMA
2.7320
0.4493
0.0026
0.0941
0.0036
0.4417
0.0439
0.3635
2.6331
0.5583
0.0077
0.0002
0.0007
0.6789
Average
MD-VA
E.PA-NJ
MD-VA
E.PA-NJ
NY
OH-W.PA-WV
IND-KY
TN-South
NYC-LI-CT-RIMA
5.2740
1.1404
0.0138
0.1098
0.0045
0.4299
0.1864
0.3377
-1.2920
0.1129
0.0039
0.0003
0.0001
-2.4698
NE
MD-VA
E.PA-NJ
MD-VA
E.PA-NJ
NY
OH-W.PA-WV
IND-KY
TN-South
NYC-LI-CT-RIMA
3.7101
0.9015
0.0016
0.0704
0.0016
0.0316
0.0645
-0.0432
2.4772
0.6739
0.0024
0.0001
0.0001
0.4334
NE
MD-VA
E.PA-NJ
MD-VA
E.PA-NJ
NY
OH-W.PA-WV
IND-KY
TN-South
NYC-LI-CT-RIMA
8.2390
2.4177
0.0124
0.0636
0.0025
0.0315
0.3741
0.0148
0.9929
0.4299
-0.0020
0.0000
-0.0001
-2.7392
Point Sources
OH-W.PAWV
0.0516
1.3029
0.5710
1.1665
2.4430
0.6007
0.0025
1.7857
0.0002
0.3099
0.0002
0.2986
2.8298
0.5555
NY
Mobile Sources
NY
OH-W.PAWV
0.0354
1.0916
0.6132
1.1448
3.4902
0.8826
0.0012
2.5714
0.0006
0.1892
0.0001
0.2268
2.3052
0.6968
Point Sources
OH-W.PAWV
0.0006
0.7401
0.1331
2.1351
2.3239
0.6983
0.0059
1.3116
0.0002
0.0088
0.0001
-0.0011
2.9909
1.3254
NY
Mobile Sources
NY
OH-W.PAWV
0.0003
0.8901
0.0299
1.6367
2.6215
1.0231
0.0021
2.3365
0.0002
0.0353
0.0005
-0.0020
2.0058
1.2002
22
IND-KY
TN-South
0.1463
0.1270
0.2132
1.3706
2.7504
0.4982
0.1248
0.1794
0.0120
0.0412
0.3025
0.9347
3.3256
0.0090
IND-KY
TN-South
0.1610
0.1992
0.4715
1.4044
3.0977
0.4153
0.1596
0.4616
0.0080
0.0360
0.2791
0.7483
10.6284
0.0058
IND-KY
TN-South
0.0413
0.2148
0.4390
2.4372
2.2100
-0.0002
0.2822
0.1932
0.0360
0.1236
0.8406
2.6808
2.5335
0.0269
IND-KY
TN-South
0.0908
0.1829
0.8299
2.0353
2.9681
-0.0034
0.2253
0.3873
0.0235
0.1079
0.7631
2.0939
6.6103
0.0179
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
MW
MD-VA
E.PA-NJ
MD-VA
E.PA-NJ
NY
OH-W.PA-WV
IND-KY
TN-South
NYC-LI-CT-RIMA
2.5916
0.4078
0.0061
0.2116
0.0090
1.2003
0.0672
1.0121
3.8519
0.4729
0.0174
0.0003
0.0020
1.7536
MW
MD-VA
E.PA-NJ
MD-VA
E.PA-NJ
NY
OH-W.PA-WV
IND-KY
TN-South
NYC-LI-CT-RIMA
2.4496
0.9301
0.0292
0.2686
0.0105
1.1772
0.1855
0.7874
-6.2273
-0.4361
0.0135
0.0002
0.0008
-3.0201
SE
MD-VA
E.PA-NJ
MD-VA
E.PA-NJ
NY
OH-W.PA-WV
IND-KY
TN-South
NYC-LI-CT-RIMA
1.8944
0.0386
0.0000
0.0002
0.0004
0.0934
-0.0001
0.1214
1.5704
0.5280
0.0032
0.0002
0.0000
-0.1504
SE
MD-VA
E.PA-NJ
MD-VA
E.PA-NJ
NY
OH-W.PA-WV
IND-KY
TN-South
NYC-LI-CT-RIMA
5.1333
0.0734
0.0000
-0.0029
0.0005
0.0809
-0.0004
0.2111
1.3585
0.3449
0.0001
0.0008
-0.0004
-1.6502
RFF 98-46
Point Sources
OH-W.PAWV
0.1082
0.6232
0.1784
0.5675
3.8936
0.9633
-0.0007
2.9821
0.0001
0.4196
0.0002
0.3563
2.1528
0.1710
NY
Mobile Sources
NY
OH-W.PAWV
0.0984
0.6255
0.3566
0.9082
5.7489
1.0913
-0.0002
2.8128
0.0003
0.3507
0.0000
0.3569
1.7938
0.4494
Point Sources
OH-W.PAWV
0.0461
2.5454
1.4014
0.7968
1.1116
0.1407
0.0022
1.0635
0.0004
0.5011
0.0002
0.5406
3.3457
0.1702
NY
Mobile Sources
NY
OH-W.PAWV
0.0076
1.7591
1.4531
0.8894
2.1004
0.5335
0.0017
2.5649
0.0014
0.1814
-0.0003
0.3256
3.1159
0.4406
23
IND-KY
TN-South
0.0232
0.0141
0.0501
0.7294
3.0427
1.2619
0.0069
0.0479
0.0000
0.0001
0.0098
0.0566
2.5379
0.0001
IND-KY
TN-South
0.0215
0.0492
0.1520
0.8243
3.7389
0.8386
0.0263
0.1011
0.0002
-0.0001
0.0124
0.0653
8.8940
-0.0001
IND-KY
TN-South
0.3746
0.1521
0.1503
0.9453
2.9984
0.2328
0.0854
0.2970
0.0000
0.0000
0.0571
0.0667
4.9055
0.0000
IND-KY
TN-South
0.3708
0.3656
0.4325
1.3537
2.5860
0.4106
0.2270
0.8963
0.0002
0.0001
0.0617
0.0855
16.3809
-0.0003
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
Table 3. Trading Ratios, by source type and episode. (MD-VA Point Sources = 1)
Episode Type
Average
NE
MW
SE
Average
NE
MW
SE
MD-VA
1
1
1
1
E.PA-NJ
1.06
0.66
1.61
0.86
NY
1.25
0.89
0.97
2.68
1.92
2.33
1.13
2.62
-0.94
-0.36
-2.04
-0.01
1.22
0.67
1.10
2.77
Point Sources
OH-W.PA-WV
1.51
1.24
1.27
2.68
Mobile Sources
1.71
1.40
1.38
3.17
IND-KY
1.01
0.92
0.79
1.68
TN-South
1.00
0.96
0.50
2.16
1.11
0.94
0.84
2.09
2.63
1.60
1.71
7.06
The most striking result for the average episode and point sources is the uniformity of
trading ratios for point sources, with the exception of the OH-W.PA-WV and, to a lesser
extent, New York. The higher ratios for these regions imply that a ton of NOx emissions
reduction from point sources in these areas is worth more (in terms of exposure reductions)
than a ton elsewhere. Given the S-R coefficients, the high effect of the OH-W.PA-WV on its
own population (coupled with high population) drives this result, in part.
This uniformity is not so evident across episodes. For instance, for the midwest
episode, reductions in E. PA-NJ are very valuable, those in the south are relatively
unproductive in reducing ozone exposures, and the OH-W.PA-WV effect is still evident.
Finally, the most striking result for mobile sources is that a ton of NOx emissions
reduction from mobile sources reduces more ozone exposures than a ton of point source NOx
emissions reductions, particularly for MD-VA. This is not surprising as the NOx emissions
from mobile sources are emitted where the people are. So a population-based exposure
measure is likely to favor mobile source reductions more than an area-weighted measure.
In terms of their trading ratios, there is greater nonuniformity across regions for
mobile sources, partly because of the scavenging effect observed for E.PA-NJ. Reductions in
the South and the Midwest are particularly productive.
D. Ozone Benefits
The population-weighted source-receptor matrices linking NOx emissions to changes in
ozone are multiplied by regional population data and estimates of the health damages per person
per unit change in ozone to estimate ancillary health benefits of NOx emissions reductions.
The estimates of health damages per person per annual change in ozone concentrations
are taken from Krupnick and Burtraw (1997), Table 4. The estimates are derived from a Monte
Carlo simulation model, including a comprehensive set of epidemiological concentrationresponse functions and values taken from the economics literature.19 The benefit estimates are
19 This model is derived from the Tracking and Analysis Framework, which was developed on behalf of the
National Acid Precipitation Assessment Program (Bloyd, 1996).
24
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
highly sensitive to the treatment of mortality risks. In line with the general reluctance of EPA
and others to ascribe mortality benefits to ozone (EPA, 1997), mortality risks are assumed to be
zero over 90 percent of the probability distribution, with 10 percent of the distribution
determined by two studies in the literature showing positive effects. The value of reducing such
risks is set at $3.2 million (rather than the consensus accidental death VSL of $4-5 million), to
account for the disproportional effects on older people and their lower than average willingnessto-pay (WTP) (as determined in Jones-Lee et al., 1985). Table 4 provides the damage per
person per unit concentration estimates used in this study (converted to 1990 dollars from 1989
dollars, to match the cost estimates) and Table 5 provides the ancillary benefit estimates applied
to each source-type per ton NOx reduced by source region. We assume that ozone benefits are
only registered during the ozone season (153 days per year). The estimates assume an average
episode, in terms of relating NOx emissions to ozone concentrations.
Table 4. Annual Unit Damages for Health Effects from Ozone
(1990 dollars/person/0.01 ppm ozone)
5%
$3.16
Mean
$7.38
95%
$86.43
These damage estimates can be interpreted in the following way. The central estimate
of $7.38 for ozone is the mean WTP of an average person for a reduction in ozone of
0.01ppm20 averaged over the year. The mean WTP is based on the expected values of the S-R
coefficients and the valuation coefficients and an aggregation algorithm designed to eliminate
double-counting of health effects.
Table 5. Monetary Benefits over the Study Domain per ton of NOx Emitted,
by Source Type and Region. Average Episode. (1989 dollars).
Point Sources
Source Regions
NY
5th percentile
Mean
95th percentile
MD-VA
61
141
1656
E. PA-N.J.
64
150
1761
75
176
2062
OH-W. PAWV
91
213
2501
IND-KY
61
143
1671
TN-South
61
141
1656
IND-KY
67
157
1839
TN-South
160
373
4368
Mobile Sources
Source Regions
5th percentile
Mean
th
95 percentile
MD-VA
116
271
3180
E. PA-N.J.
-57
-133
-1563
NY
74
172
2020
OH-W. PAWV
104
242
2833
20 The current daily peak 1-hour standard is 0.12 ppm, the new 8-hour averaged standard is 0.08 ppm.
25
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
VI. RESULTS OF THE POLICY ANALYSIS
Here we present the results of the three major policy analyses outlined in the
Introduction. First, we examine the outcomes under the planned Clean Air Act and OTAG
regulatory (CAC) programs. We refer to the Clean Air Act scenario as the CAA scenario and
to the additional OTAG controls as the OTAG scenario. In the second part of the analysis, we
compare the CAC policies to various hypothetical least-cost outcomes as a guide to future
policy decisions. Finally, we examine possible NOx trading rules for meeting future ozone
goals, and their impact on the Bay.
Policy Analysis 1: The Impact of Clean Air Act Regulations and OTAG Controls
Baseline Emissions
Tables 6 and 7 below show the forecast baseline or initial levels of NOx emissions and
Bay nitrate loads resulting from air sources before either the CAA or OTAG regulatory
scenarios are implemented. All forecasts are for annual activity in the year 2005. Ohio and
then Pennsylvania contribute the largest share of NOx emissions, but Pennsylvania emissions
have by far the largest impact on Bay nitrate loadings, with Maryland, Virginia, Ohio, New
York and West Virginia emissions also contributing significantly to nitrate loadings.
Table 7 shows how the different NOx emissions sources account for both total NOx
emissions and the impact on nitrate loads in the baseline case. In the Bay jurisdictions alone,
mobile sources and utilities account for approximately equal shares of NOx emissions, while
in the Airshed, utilities account for substantially more than mobile sources. Nitrate loadings
are proportionate to emissions by source category in the Bay states and mobile sources
contribute proportionately more to Bay nitrate loadings than their share of emissions. (These
results parallel those of the Pechan Analysis (1996), though there are some differences
because the emissions inventories have been updated).
Impact of the Clean Air Act Scenario
The results of the Clean Air Act requirements are shown in the top section of both
Tables 8 and 9. Table 8 shows that reductions in NOx emissions due to controls under the
Clean Air Act result in reductions in nitrate loadings to the Bay of 17.5 million pounds a year
from the airshed -- about 27 percent of the airborne nitrate load -- with the Bay jurisdictions
accounting for about half of the total. Although utilities are responsible for about half of the
baseline airborne nitrate loadings in the airshed (Table 7 above), they account for almost 80
percent of the loadings reductions in the region. Under the Clean Air Act controls, the
percentages are a little lower for the Bay jurisdictions alone, but the finding that utilities are
required to reduce emissions by a relatively large percentage holds for the smaller Bay region
as well. Examination of NOx emissions reductions in both regions (Table 8) makes it clear
that utilities have been heavily targeted for NOx emissions reductions under Clean Air Act
mandates.
26
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
Table 6. No-Control Baseline NOx Emissions and Nitrate Loads. Year 2005
State
Bay jurisdictions
DC
Maryland
Pennsylvania
Virginia
Bay total
Other Airshed States
Delaware
Indiana
Kentucky
Michigan
New Jersey
New York
North Carolina
Ohio
Tennessee
West Virginia
Total Airshed
Emissions
(thousand
tons/year)
% of total
emissions
Nitrate Load
(thousand lbs/year) % of total load
13
283
781
380
0.2%
5.4%
14.9%
7.2%
27.7%
281
9,145
15,459
7,810
0.4%
13.8%
23.4%
11.8%
49.4%
63
350
444
353
250
379
339
985
167
455
5,242
1.2%
6.7%
8.5%
6.7%
4.8%
7.2%
6.5%
18.8%
3.2%
8.7%
2,047
1,546
1,962
2,226
1,224
6,719
2,569
7,937
933
6,328
66,186
3.1%
2.3%
3.0%
3.4%
1.8%
10.2%
3.9%
12.0%
1.4%
9.6%
Table 7. Share of NOx Emissions and Nitrate Load Reductions by
Source Category. No Control Baseline
Mobile Sources
Utilities
Non-Utilities
Bay jurisdictions
% nitrate
% emissions
loadings by
by source
source
45%
46%
44%
44%
11%
10%
27
Airshed
% emissions by
source
37%
53%
10%
% nitrate loadings
by source
40%
50%
10%
Scenario
CAA
Mobile Sources
Utility
Non-Utility
OTAG
Mobile Sources
Utility
Non-Utility
Table 8 - Annual Nitrate Load and NOx Emissions Reductions, Benefits and Costs ($1990) . Year 2005
Bay jurisdictions
Airshed
NOx
Mean
Mean
Nitrate
Cost
High
NOx
Nitrate
Cost
Emissions
Load
(M$)
Benefit benefits
Emissions
Load
(M$)
Benefits
Reductions Reductions
(M$)
(M$)
Reductions Reductions
(M$)
(000 tons)
(000 lbs)
(000 tons)
(000 lbs)
357
8,099
417
33
387
1,462
17,543
1,185
125
15%
17%
244
6
67
9%
11%
513
7
72%
71%
67
24
279
83%
79%
423
108
13%
12%
105
3
41
8%
10%
248
10
310
13%
85%
2%
7,049
13%
85%
2%
517
96
356
65
26
2
23
1
307
25
276
6
1,024
9%
84%
7%
13,669
10%
85%
5%
Table 9 - Gross and Net Costs per pound of Nitrate Reduced
Bay jurisdictions
1,725
286
1,114
326
High
Benefits
(M$)
84
4
75
5
1,468
88
1,265
115
985
46
876
63
Airshed
Scenario
CAA
Mobile Sources
Utility
Non-Utility
Gross
Cost/ton of
NOx
reduced
1,168
4,436
262
2,283
Gross
Cost/lb of
nitrate
reduced
51
126
5
57
Cost/lb net
of mean
ozone
benefits
47
123
3
55
Cost/lb net
of high
ozone
benefits
4
91
-15
34
Gross
Cost/ton of
NOx
reduced
811
4,039
349
2,000
Gross
Cost/lb of
nitrate
reduced
67
265
31
136
Cost/lb net
of mean
ozone
benefits
60
261
23
130
Cost/lb net
of high
ozone
benefits
-16
220
-61
73
OTAG
Mobile Sources
Utility
Non-Utility
1,668
2,462
1,348
9,286
73
102
60
435
70
100
56
432
30
76
13
393
1,685
2,979
1,298
4,657
126
206
96
465
120
203
90
458
54
172
21
376
28
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
The costs of obtaining reductions in nitrate loadings resulting from regulations that are
going to be undertaken to improve air quality can be considered as "free" to the Bay. These
reductions would occur at no additional cost to the Bay. It is nonetheless interesting to know
what the costs of these air reductions are. The top part of Table 8 shows the costs for the
Clean Air Act reductions. These costs are estimated to be about $400 million for the Bay
jurisdictions annually and $1.185 billion annually for the airshed by the year 2005. Mobile
source reductions appear to have the highest costs in both the Bay jurisdictions and the entire
airshed. Utilities are the most cost-effective source of NOx emissions reductions, with low
costs and the greatest emission reductions.
The benefits of improvements in ozone resulting from NOx emissions reductions for
the different regions are also shown at the top of Table 8 for the CAA scenario.21 The mean
value of benefits is small, less than 10 percent of the costs in the Bay and a slightly higher
percentage of costs in the airshed. However, the "high" estimate of benefits is greater than the
costs, meaning that net costs are negative in both the Bay and the airshed. The high estimate
of benefits is much higher than the mean benefit estimate because it includes potential
mortality effects from NOx emissions. The different benefits estimates are discussed at
length in section V.D.
Table 9 examines the cost-effectiveness of the CAA controls from both the Bay states
and the entire airshed. The first columns for each region attribute all of the costs first to NOx
emissions reductions and then to reduction of nitrate loadings to the Bay. The Bay
jurisdictions are able to more cost-effectively control NOx to reduce nitrate loadings to the
Bay compared to the entire airshed. The cost-effectiveness is $51/pound for the Bay
jurisdictions and $67/pound for the region as a whole. However, it is not the case that Bay
jurisdiction's control of NOx emissions are more cost-effective than those in the airshed as a
whole. The Bay states reduce NOx emissions at an average cost per ton of $1,168 compared
to only $810 for the airshed. Rather, the proximity of the Bay states to the Bay makes them
more cost-effectiveness in reducing nitrate loading.
In terms of sources, Table 9 shows that in both the Bay region and the airshed, the
gross costs per pound of nitrate reduced are dramatically lower for utilities than for the mobile
sources under CAA controls. It makes sense, then, that the Clean Air Act targeted utilities so
strongly. It is also noteworthy that the cost per ton of NOx reduced for utilities is much lower
for the Bay jurisdictions than for the airshed as a whole.
Table 9 also shows cost-effectiveness results when costs are attributed to both the
reduction of nitrate loads in the Bay and ambient ozone levels. The benefits of ancillary
ozone improvements are netted out of the costs before the cost per ton is calculated. "High"
and "mid" measures of benefits are used. As Table 8 shows, the mean benefits are quite small
and have little impact on the cost-effectiveness results. However, again, the high estimate of
benefits, which includes potential mortality effects of NOx emissions, has a large impact on
the cost-effectiveness results. The "high" benefits outweigh the costs for the airshed region as
21 See discussion of ozone benefit estimation in Section V.
29
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
a whole and costs per ton are negative $16. The reason for this seems to be that the benefits
from the utility sector are large enough to outweigh total costs for the region.
Another important point about the ozone benefits is that ton for ton, mobile sources
result in lower benefits than do utilities: $59/ton for mobile sources versus $89/ton for
utilities. This occurs in spite of generally higher source-receptor coefficients for mobile
sources (which emit NOx where the people are) primarily because emission reductions from
mobile sources in E.PA-NJ have a strong ozone-enhancing effect (termed NOx scavenging).22
If the benefit per ton estimates are examined for a particular state that does not have
scavenging (such as Maryland), mobile sources are far more productive in creating health
benefits: $161/ton versus $78/ton for mobile and utility sources, respectively.
Impact of the additional OTAG controls
The lower sections of Tables 8 and 9 show the results of additional controls being
considered for controlling NOx emissions as part of the OTAG scenario. Under this scenario,
nitrate loadings fall by 13.7 million pounds in the entire airshed -- or another 35 percent from
the 48.5 million pound CAA level -- and NOx emissions are reduced by one million tons
annually. About half of these improvements in nitrate reductions come from Bay jurisdictions,
but this Bay region accounts for only about one third of the NOx emissions. As under the
CAA scenario above, NOx emissions reductions in the Bay jurisdictions have a larger impact
on nitrate loads to the Bay than NOx emissions reductions in other parts of the airshed.
Even more of the reductions under the OTAG scenario come through utilities
(85 percent) compared with the Clean Air Act controls (Table 8). And, mobile sources make
up a smaller share in both the Bay jurisdictions and in the region as a whole: 13 percent in the
Bay jurisdictions and 10 percent in the airshed, compared to 17 percent and 11 percent under
the Clean Air Act controls.23
However, from Table 8, it is clear that the costs of control change from the CAA case.
Emission reductions are lower and the costs of achieving them are significantly higher under
the OTAG case; in other words, the marginal costs of greater reductions rise with tighter
controls, as we would expect. There are also some changes in the sources of reductions from
the base to CAA case. The costs of controlling utility sources under OTAG are much higher,
and the mobile source controls are lower. The latter occurs because of the assumptions
within the Pechan data that the additional costs of LEV vehicles will be relatively low.24
Also, note that the ozone benefits are smaller for both mean and high benefits because NOx
emissions reductions are smaller than they were under the CAA scenario.
Table 9 shows again what we would expect for the stricter controls in the OTAG
scenario relative to the CAA scenario -- that the cost-effectiveness of NOx emissions
22 See Table 2 above for S-R coefficients.
23 Recall that under the OTAG scenario, mobile source controls include only the introduction of Low Emission
vehicles.
24 Note that there are no LEVs in the CAA scenario but that LEVs are mandated in the OTAG scenario.
30
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
reductions worsens for the OTAG scenario, climbing for the airshed from $811/ton under the
CAA scenario to $1,685/ton for the OTAG scenario. For the Bay jurisdictions, whose costs
were quite a bit higher than those for the airshed as a whole in the CAA scenario, costs
increase less dramatically, to the point where average NOX cost-effectiveness is no different
for the Bay and the region ($1,668/ton for the Bay and $1,685/ton for the region).
From the perspective of the Bay community, it is cost-effectiveness in terms of nitrate
loading reductions that is most important. These results are shown in Table 9. Costs for the
airshed also climb in terms of this measure under the OTAG scenario (from $67/pound for the
CAA scenario to $126/pound for the OTAG scenario). But the Bay jurisdictions widen their
advantage as a cost-effective source of nitrate loading reductions, being 42 percent cheaper
per pound reduced versus 24 percent cheaper per pound in the CAA scenario. Thus, the Bay
jurisdictions are an attractive target for further airborne nitrate loading reductions.
Which sources within the Bay jurisdictions are the most attractive targets? For the
CAA scenario, utilities were the most cost-effective (using gross costs), at $5/pound, followed
by non-utility point sources at $57/pound, and mobile sources at $ 126/pound. For the OTAG
scenario, there are some dramatic changes. Utilities are still the most cost-effective source, at
$60/pound, but mobile source costs actually fall to $102/pound. The decrease in costs for
mobile sources is, as we discussed above, primarily because LEV vehicles are assumed to
have relatively low cost, at $76 per vehicle, and because emissions reductions from LEVs are
large especially in conjunction with enhanced I&M (see Appendix 3). Further regulation of
non-utility point sources in the Bay jurisdictions appears not to be cost-effective, at
$435/pound reduction.
Tables 10 and 11 show the impact of the CAA and OTAG scenarios across states, for
both Bay jurisdictions and airshed states. Table 10 shows percentage reductions in both NOx
emissions and nitrate loads to the Bay for each state that will result from successively more
stringent regulations. The first column shows the percentage change in NOx from the
baseline to the CAA, and the second column show the percent change from the CAA scenario
to the OTAG scenario. The percentage variation differs by state, and some states doing
relatively more under the OTAG scenario, and some less. The two states with the greatest
baseline amount of NOx, Pennsylvania and Ohio, will reduce by a relatively large percentage,
but some states will reduce by even higher percentages. Looking at the fifth and sixth
columns of Table 10 we see that the percentage reductions in nitrate loadings are roughly
similar to the percentage reduction in NOx emissions reductions.
However, the share of emission reductions among states does seem to change over
time and vary with the pollutant under examination. Under the OTAG scenario, the share of
NOx emissions reductions required for the Bay jurisdictions rises, and only a few of the other
airshed states increase their share of the total reductions compared to the CAA case (the share
of reductions for the Bay jurisdictions rises from 24 percent to 31 percent from the CAA to
OTAG cases).
31
Table 10. State Contributions of NOx Emissions and Nitrate Loading to the Bay from Air Sources
NOx Emissions Reductions
Nitrate Loading Reduction
State
Percent reduction in
NOx Emissions (%)
Bay Jurisdictions
DC
Maryland
Pennsylvania
Virginia
Total
Other Airshed States
Delaware
Indiana
Kentucky
Michigan
New Jersey
New York
North Carolina
Ohio
Tennessee
West Virginia
Total Airshed
State Share of total NOx
Emissions Reductions
Percent reduction in
Nitrate Loadings (%)
State share of total Nitrate
Loading Reduction
CAA
23
27
30
11
OTAG
20
33
32
20
CAA
0
5
16
3
24
OTAG
0
7
17
7
31
CAA
23
28
30
10
OTAG
19
34
32
18
CAA
0
15
27
5
47
OTAG
0
17
25
9
51
27
33
34
31
27
17
21
32
17
36
30
22
35
16
16
21
25
24
40
44
1
8
10
7
5
4
5
22
2
11
100%
1
5
10
4
3
7
7
16
5
13
100%
28
33
34
31
27
17
21
32
17
36
32
22
35
16
17
21
26
24
40
44
3
3
4
4
2
7
3
14
1
13
100%
3
2
3
2
1
9
4
9
2
13
100%
32
Table 11. Costs of NOx Emissions and Nitrate Loading Reduction
(Assume gross costs unless stated otherwise)
State
Bay Jurisdictions
DC
Maryland
Pennsylvania
Virginia
Other Airshed States
Delaware
Indiana
Kentucky
Michigan
New Jersey
New York
North Carolina
Ohio
Tennessee
West Virginia
Total Airshed
Cost of NOx
Emissions Reductions
(millions of dollars)
CAA
OTAG
Cost-Effectiveness of Nitrate
Loading Reductions
CAA
$/lb
$/lb
OTAG
$net/lb
(high benefits)
Cost Effectiveness of
NOx Emissions
Reductions ($/ton)
CAA
OTAG
12
123
190
93
5
96
301
115
174
48
40
117
122
43
87
89
73
14
35
43
3,833
1,596
800
2,323
2,500
1,433
1,728
1,721
21
90
63
19
184
98
28
214
12
40
1,187
37
97
151
86
57
109
121
258
105
188
1,726
36
177
94
27
549
84
52
84
76
17
1,575
80
421
337
342
388
91
234
201
340
106
2,881
54
251
143
163
408
38
131
59
202
19
1,633
1,224
785
416
171
2,703
1,506
400
678
425
241
17,101
2,657
1,871
1,495
2,148
1,900
1,627
1,782
1,620
1,900
1,471
25,853
33
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
However, the Bay states contribute a much greater share of the nitrate loading reductions and
this share increases from 47 percent to 51 percent under the OTAG scenario. This is to be
expected since the Bay states are adjacent to the Bay; but it can also be looked at the other
way, that about half of the air contributions come from states that are not directly adjacent to
the Bay, such as New York, Ohio, and West Virginia. And, there are also some important
differences in how total nitrate loading reductions are distributed across states under the two
scenarios. For example Pennsylvania contributes almost 3 times as much nitrate loading
reduction in the OTAG scenario as Virginia, even though their initial nitrate loadings to the
Bay were only about double Virginia's.
Table 11 shows the total costs of NOx emissions reductions and two costeffectiveness measures by state. As we would expect, total costs are the highest for the states
which have the largest emissions, or which are making the most reductions, such as
Pennsylvania, Ohio and West Virginia. However, what is most interesting in Table 11 is that
the gross costs per pound of nitrate load reduced are much lower in the Bay jurisdictions
(except D.C.) than most of the other states. In the OTAG case, Maryland's cost per pound
nitrate load reduced ($43) are much lower than for any other state -- about half that of the
other Bay states. Costs per pound reduced are usually higher in the OTAG case, but not
always -- Virginia and New Jersey actually show costs per pound falling. The costs per ton of
NOx reduced vary a great deal across states. The variation is less under OTAG and costs are
higher everywhere except in Virginia and D.C.
Finally, including "high" benefits in the cost-effectiveness calculation results in a large
variation in cost/pound nitrate loadings reductions across the states. The net costs per pound for
New Jersey sources is more than 10 times the costs in the Bay jurisdictions. However, the
relative distribution of cost-effectiveness between states is not changed much. Although the
cost-effectiveness estimates including the high ozone benefits are much lower, the relative
ranking of the states does not appear to change much. This is because the source-receptor
relationships linking NOx emissions to ozone changes over the airshed are similar across states.
Policy Analysis 2: Comparison of the CAC Scenarios to Scenarios for Achieving Nitrate
Loading Reductions
In this section we examine results for the entire airshed and then present details of the
changes in technology choices for the various scenarios, and the effect of different ozone
episodes on the least-cost results when ozone benefits are included in the optimization. Then,
we turn to results for runs with the Bay jurisdictions alone.
Policy Analysis 2.a: Airshed Analysis
In this section, we examine a series of potential least cost control policies that, as the
name suggests, achieve some target emissions reduction at the lowest possible cost. Suppose
that, like those interested in achieving the goals of the Clean Air Act, the Chesapeake Bay
community also has a claim on NOx emissions reductions. The Bay policy community might
34
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
want to focus states' efforts entirely on cleaning up the Bay and, therefore, would want to
know the least cost way of getting nitrate loading reductions and compare these to the OTAG
scenario. These least cost results can provide information about how policies to reduce nitrate
loads to the Bay can be better targeted.
There are a number of different ways of looking at possible least-cost ways of
reducing nitrate loadings to the Bay, all of which result in nitrate loading reductions
equivalent to that from the OTAG scenario examined above. With the controls mandated by
that scenario, nitrate loads to the Bay are forecast to fall by 13.67 million pounds a year
(Table 8). We compare this OTAG scenario to three least-cost scenarios. This analysis
minimizes the costs of reducing nitrate loadings across all of the states in the airshed. We
examine the situation where the controls are limited only to the Bay jurisdictions in a separate
section below.
The three least cost scenarios are:
Scenario i. Gross Least cost. The least cost way of achieving the nitrate loading
reduction goal, minimizing gross abatement costs.
Scenario ii. Net Least cost; minimizing net costs instead of gross costs. This scenario
is like (i). above but seeks a NOx emissions reduction allocation that meets the nitrate
goal cost-effectively in light of the additional objective of maximizing health benefits
from ozone reductions.
Scenario iii. Emission trading. The least cost solution for meeting the nitrate loading
target may be too complex to implement because it requires each air source to trade
emissions at different ratios with every other source.25 As an alternative we examine a
different incentive scenario, called emissions trading, where emissions are exchanged or
traded ton for ton. It embodies the idea that a "least-cost" allocation of NOx abatement
activity would take place, but using emissions rather than loadings as a unit of account.
Such allocations are altered until the same nitrate loading reduction goal is met.
Scenario i: Least cost
Table 12 compares the least cost scenario for getting 13.7 million pounds of nitrate
loading reductions to the OTAG scenario. The least cost allocation of reductions (minimizing
gross costs, section I of Table 12) would attain this target at only one-third the cost of the
OTAG scenario allocation, while resulting in NOx emissions reductions that are 80 percent of
those under the OTAG allocation. This cost-savings arises out of differences in both the
geographic and source-type allocations of NOx emissions reductions.
25 The controversy over such trading in the OTAG debate is illustrative of the implementation difficulties for
this approach.
35
Table 12. Summary Results for OTAG and Least Cost Scenarios
Total Nitrate Reduction = 13.67 million lbs/yr for all Scenarios
Total
OTAG
Share of Nitrate Loading Reduction
(% of total)
NOx Emissions reductions
(thousands of tons)
Cost ($millions)
Cost/pound nitrate loadings reduced
Net cost/pound nitrate loadings reduced
(mean benefits)
I. Least Cost for Nitrate Loading Goal
(minimize gross costs)
Share of Nitrate Loading Reduction
(% of total)
NOx Emissions reductions
(thousands of tons)
Cost ($millions)
Cost/pound nitrate loadings reduced
Shadow Price ($/lb Nitrates)
II. Least Cost for Nitrate Loading Goal
(minimize net costs, high benefits)
Share of Nitrate Loading Reduction
(% of total)
NOx Emissions reductions
(thousands of tons)
Cost ($millions)
Ozone Benefits ($millions)
Net Costs ($millions)
Net cost/pound nitrate loading reduced
(high benefits)
Shadow Price ($/lb Nitrates)
III. Emissions Trading
(minimize net costs, high benefits)
Share of Nitrate Loading Reduction
(% of total)
NOx Emissions reductions
(thousands of tons)
Cost ($millions)
Net Costs ($millions)
Net cost/pound reduced
(high benefits)
Shadow Price ($/ton emissions)
Regions
Other
Bay juris- Airshed
dictions
States
Sources
Mobile
Sources
Utilities
Nonutility
Point
Sources
100%
51%
49%
10%
85%
5%
1,024
310
714
96
858
70
$1,725
$126
$120
$517
$73
$1,208
$182
$286
$206
$1,114
$96
$326
$465
100%
62%
38%
26%
66%
8%
813
374
439
186
543
84
$581
$43
$111
$371
$44
$210
$40
$90
$26
$440
$49
$51
$47
100%
57%
43%
23%
69%
8%
875
343
532
180
608
87
$655
$921
-$266
-$19
$331
$377
$324
$543
$94
$193
$504
$640
$56
$89
-$8
-$40
-$31
-$14
-$31
100%
53%
47%
24%
68%
8%
988
324
663
214
680
93
$854
-$219
-$16
$283
$571
$189
$601
$63
-$11
-$22
-$25
-$12
-$28
$44
$786
36
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
Not surprisingly, the Bay jurisdictions take a larger share of the emission reductions in
the least-cost scenario, 46 percent compared to 30 percent in the OTAG scenario (see
Figure 4). In addition, the mobile source share in the airshed rises dramatically, from
10 percent in the OTAG regulatory scenario up to 26 percent in the least-cost scenario. And,
although the breakdown by source type for the Bay states are not shown, the mobile source
share goes from 12 percent in the OTAG scenario to 28 percent in the least cost scenario.
These increases are at the expense of utility reductions. The shift away from utilities toward
mobile sources may seem unlikely given the results presented earlier that the cost per pound
of nitrate removed were much higher, on average, for mobile sources under the OTAG
scenario. It must be, therefore, that although the average costs are high for mobile sources in
the OTAG scenario, there are some mobile source options that, at the margin, are quite low.
For instance, the average cost of reducing a pound of nitrate from mobile sources is $23, but
the marginal cost of adding a LEV program to an existing high-enhanced I&M program is
only $1/lb nitrate removed.26
Average costs per pound of nitrate reduced falls under the least cost scenario to $43/lb
from $126/lb with OTAG. The distribution of the costs per pound reduced among states is
shown in Figure 5. Costs per pound are actually negative in three states (even before
accounting for ancillary benefits): Kentucky, the District of Columbia and New Jersey.
Because a least-cost strategy involves fewer emission reductions than the OTAG
strategy, it might be that there would be lower ozone benefits, thereby canceling out any cost
savings. In fact, ozone benefits are so low in the mean-benefits estimate that this
consideration is not an issue. We find that the mean-benefits in the least-cost scenario are $65
million, compared to $84 million in the command and control scenario, making net costs $516
in the least-cost scenario compared to $1,642 in the command and control scenario (not
shown). The least cost scenario still yields net costs that are less than one-third of those for
the command and control scenario. This implies, for Bay policy making, that from both an
ozone and a Bay perspective there can be substantial savings from targeting low cost sources
of nitrate reductions.
Another way of gauging the costs of this scenario is by examining the shadow price of
nitrate reductions, which is $111 per pound. This means that the last unit of nitrates reduced
to just meet the nitrate loadings reduction target costs $111/pound. Any further tightening of
the nitrate loadings reductions target will cost at least this much per pound. This figure can be
compared to the average cost, which is far lower at $23/pound.
Scenario ii: Net Least cost; minimizing the net costs of nitrate loading reduction
Table 12 shows the results for a stronger test of the impact of including ozone benefits
in the analysis. Part II examines the scenario where the allocation is designed to minimize the
26 This extreme example is due to assumptions made by Pechan at the direction of EPA. Maximum LEV credits
are only assigned with high enhanced I&M programs. As a result, extremely high benefits, in terms of NOx and
VOC reductions, are assigned when this coupling occurs (see Appendix 3 for more discussion of this issue).
37
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
38
RFF 98-46
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
39
RFF 98-46
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
costs of reducing nitrate loads with the ozone benefits netted out. With the mean-benefits
assumption, net costs are virtually unchanged when such an allocation is made (at $516
million) but there is some minor rearrangement of abatement (this run is not included in the
table since there is so little impact).
To gain further insight, we redid the allocation using our "high" estimate unit value for
ozone benefits. These results are shown in part II of Table 12. In this case, net costs are
negative: $296 million in "net" benefits, with abatement costs of $655 million and health
benefits of $921 million. Emission reductions are larger than in the original least-cost
scenario: 875,000 tons versus 813,000 tons.
What about the differences in the allocation of abatement activity? Geographically,
optimizing on net benefits instead of gross costs reduces the Bay jurisdictions' contributions
to NOx emissions reductions for all source categories -- a 13 percent reduction for mobile
sources and 6 percent and 7 percent reductions for utilities and point sources in the Bay,
respectively. The lion's share of these differences can be accounted for by Pennsylvania,
where, because of scavenging, NOx emissions reductions from mobile sources are found to
increase ozone. Thus, to maximize ozone benefits requires lower reductions (or even
increases) in NOx emissions from mobile sources in this state. (See Figures 6 and 7 for stateby-state details). For the airshed, NOx emissions reductions are 10 percent higher than they
are in the gross cost scenario, because ozone reduction benefits are internalized in abatement
option choices. These added emission reductions are primarily those from utilities in non-Bay jurisdictions. Mobile source emission reductions even fall slightly relative to the gross
cost scenario.
Finally, in Table 12 section II, note that the shadow price of nitrate reductions in this
high benefit case is only $44/pound, much lower than the $111 shadow price when only gross
costs are considered. This at first appears contrary to expectation because this scenario
achieves greater NOx emission reduction which is likely to result in higher marginal costs.
However, the associated marginal ozone benefits more than offset these costs.
Scenario iii: Emissions trading
This scenario achieves the target level of nitrate loading reductions but does so by
allowing ton for ton trading among all NOx sources. Sources will therefore trade until the
cost of removing a ton of NOx is equal among all sources. Section III of Table 12 shows
results for the emissions trading scenario in which the high estimate of benefits is netted out
from costs. Emission trading is not the lowest cost way of meeting the nitrate loading
reduction goal (net costs are higher than those in the least-cost scenario (-$219 versus -$266
million), but still far lower than for the OTAG regulatory scenario ($740 million is costs of
this scenario net of high benefits, see Table 8 above). Emissions trading is likely to be a more
feasible way of achieving a lower cost solution, since tons of NOx can be traded one for one,
making the transactions costs low. Under the true least cost scenario, sources might have to
trade at different ratios, depending on the region and source type.
40
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
41
RFF 98-46
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
42
RFF 98-46
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
The costs of the emissions trading scenario are higher than the least-cost scenario
because there are more emission reductions: 988 million tons versus only 875 million tons in
the least-cost scenario (high benefits). Who must undertake the added emission reductions?
It is not sources in the Bay jurisdictions, whose share of NOx emissions reductions falls from
39 percent of the total in the least-cost scenario to 33 percent of the total in the emissions
trading scenario. In fact, most non-Bay jurisdiction increase their emissions and every Bay
state decreases them. In fact, some Midwestern states like Tennessee and Michigan more
than double their emission reductions of NOx under the emission trading scenario which
includes high benefits.
In summary, the emission trading scenario may be more politically feasible, but is
18 percent more expensive than the least cost scenario, and shifts control away form the Bay
states to the Midwest.
Technology Choices by Scenario
The above narrative has examined the effects of alternative scenarios on costs and
NOx emissions reductions (and other factors) amongst the states and across source types.
However, we have not yet discussed how technology choices change within any given source
type between least cost and regulatory scenarios.
As shown in Table 13, turning first to mobile source details, the command and control
scenario for the CAA leaves thirty-eight percent of the 84 mobile source regions without any
mobile source emissions-reducing program. Low-enhanced I&M (20 percent) and highenhanced I&M with reformulated gasoline (17.9 percent) are the most common CAA
programs undertaken. These technology forecasts for mobile sources are based on state by
state polling of what technologies were to be put in place as part of implementation of the
CAA in each jurisdiction. Our analysis uses these full responses as the OTAG scenario, and
the CAA choices for mobile sources were defined as the OTAG selections minus LEV
programs. As a result, only the OTAG scenario includes LEV vehicle options. In the Pechan
database, the LEV option can be used in only the Ozone Transport Region (OTR) or only
Massachusetts and New York, or in the entire region. When LEV vehicles are chosen in
combination with High Enhanced I/M very large emission reductions result (see Appendix 3
for more detail on this). Any of the technology options can be chosen under the least cost
scenario, but the result is that there is much less variation in programs selected. 27 percent of
regions do not undertake any emission reducing program, but all regions that do reduce
emissions establish high-enhanced I&M or high- enhanced I&M with LEV.
43
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
Table 13. Most common technologies in mobile source regions
(n=84)
CAA
No policies
Low Enhanced I/M
High Enhanced I/M, with reformulated gas
High Enhanced I/M, without reformulated gas
Basic I/M, with reformulated gas
Low Enhanced I/M, with reformulated gas
38.1%
20.2%
17.9%
7.1%
4.8%
4.8%
92.9%
OTAG
Min LEV, OTR states
Low Enhanced I/M, and Min LEV OTR states
High Enhanced I/M, reformulated gas, Max LEV OTR states
Min LEV, outside OTR states
High Enhanced I/M, Max LEV outside OTR
Basic I/M, and Min LEV outside OTR states
26.2%
20.3%
17.9%
11.9%
6.0%
3.6%
85.9%
Least Cost
High Enhanced I/M, Max LEV
No policy
High Enhanced I/M
69.0%
27.4%
3.6%
100%
44
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
Considering details for the electric utility sector (Table 14), coal/wall boiler with lowNOx burner (LNB) is the most common technology (26.5 percent) used by the 656 utilities in
the CAA command and control scenario. In the OTAG command and control scenario, utilities
are most likely to use coal/wall boilers with selective non-catalytic reduction + selective
catalytic reduction (SNCR+SCR) (23.9 percent) and coal/tangential boilers with SNCR+SCR
(22.4 percent). In the least-cost allocation scenario, coal/wall boilers with SNCR+SCR are used
less frequently (7.5 percent). Instead, the most common technology selected is coal/wall boiler
with LNB+ overfire air (OFA) (18 percent). This technology is probably preferred because the
addition of OFA given the existence of LNB may be more efficient than a switch to a different
set of technologies, such as SNCR+SCR.
The most common technologies used by the 4,017 point sources are the same in the
CAA and OTAG command and control and the least-cost scenarios. Industrial, Commercial,
and Institutional (ICI) Boilers-residual oil with LNB and ICI boilers-coal/stoker with SNCRUrea account for 32-38 percent of the technologies used in each scenario. A reason for the
similarity between technologies used in the CAA and least-cost scenarios is that the cost
minimization solution results in only 22 percent of point sources (as opposed to 55 percent of
utilities) adopting a technology different from the CAA technology. Of the sources using
technologies that differ from the CAA, ICI Boilers Natural Gas with OT+WI and Internal
Combustion Engines-Gas with L-E are the most common technologies used.
The Impact of Alternative Ozone Episodes on Model Results
Three different ozone episodes (representing meteorological conditions during "typical"
but not extreme high ozone conditions in the Northeast, in the Midwest, and in the Southeast)
were specified (see section V.C). Table 15 shows the effects of different ozone episodes on the
results for the least cost scenario that minimizes the costs net of high ozone benefits of
achieving the 13.7 million pounds nitrate loading reduction goal. The Midwest and Southeast
episodes are quite similar to each other and are somewhat different from the average of all
episodes. They require lower NOx emissions reductions to meet the nitrate loadings target,
which implies that the target will be met at a lower cost. The Northeast episode is quite
different because reductions in NOx during this episode are so productive in reducing ozone,
NOx reductions and costs are largest for these "NE" conditions.
For many regions, controlling utilities is more cost effective during the NE episode, so
there is a shift toward more NOx emissions reduction from utilities and less reduction from
mobile sources under this episode. This appears to push utility costs up more than it pushes
mobile source costs down.
45
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
Table 14. The Most Common Technologies in CAA, OTAG and Least Cost
(OTAG) Scenarios. Utilities and Point Sources
Five most common technologies for utilities (n=656)
CAA
Utility Boiler-Coal/Wall: LNB
Utility Boiler-Coal Tangential: LNC1
Gas Turbines-Natural Gas: None
Utility Boiler-Oil-Gas/Wall: LEA+BOOS
Utility Boiler-Oil/Gas/Tangential: None
26.5%
21.5%
9.0%
7.6%
5.8%
70.4%
OTAG
Utility Boiler-Coal/Wall: SNCR+SCR
Utility Boiler-Coal/Tangential SNCR+SCR
Gas Turbines-Natural Gas: None
Utility Boiler-Oil-Gas/Wall: LNB+SNCR
Utility Boiler -Coal/Vertical: Combustion Control
23.9%
22.4%
9.0%
6.0%
5.5%
66.8%
Least Cost for OTAG
Utility Boiler-Coal/Wall: LNB+OFA
Utility Boiler-Coal/Tangential LNC1
Utility Boiler-Coal/Wall: SNCR+SCR
Gas Turbines-Natural Gas: None
Utility Boiler-Coal/Wall: LNB
18.0%
12.4%
7.5%
7.0%
6.1%
50.9%
Five most common technologies for point sources (n=4,017)
CAA
ICI Boilers-Residual Oil: LNB
ICI Boilers-Coal/Stoker: SNCR-Urea
ICI Boilers-Natural Gas: LNB
ICI Boilers-Coal/Stoker: None
Internal Combustion Engines-Gas: L-E
18.2%
14.7%
8.5%
8.1%
6.3%
55.8%
OTAG
ICI Boilers-Residual Oil: LNB
ICI Boilers-Coal/Stoker: SNCR-Urea
ICI Boilers-Coal/Stoker: None
ICI Boilers-Natural Gas: LNB
Internal Combustion Engine-Gas: L-E (med speed)
17.1%
15.1%
7.8%
7.0%
6.2%
53.2%
Least Cost for OTAG
ICI Boilers-Residual Oil:LNB
ICI Boilers-Coal/Stoker: SNCR-Urea
Internal Combustion Engine-Gas: L-E
ICI Boilers-Coal/Stoker: None
ICI Boilers-Natural Gas: LNB
21.6%
16.4%
10.3%
6.5%
5.7%
60.5%
46
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
Table 15. Effects of Different Ozone Episodes on Results From Net
Least Cost Scenarioa
Least Cost for Nitrate Goal
(minimize net costs, high benefits)
NOx Emissions reductions
(millions of tons)
Gross Costs ($millions)
Ozone Benefits ($millions)
Net cost/pound nitrate loading reduced
(high benefits)
Shadow price ($/lb nitrates)
a.
Average of
all episodes
875
Northeast
episode
900
Midwest
episode
857
Southeast
episode
865
$655
$921
-$19
$701
$1,093
-$29
$645
$834
-$14
$666
$802
-$10
$44
$47
$55
$59
Minimize costs net of high ozone benefits to achieve 13.7 million pounds nitrate loading reduction goal.
Policy Analysis 2.b: Bay Jurisdictions Analysis
Another way to gain insight into policies the Chesapeake Bay jurisdictions might
favor is to suppose that no NOx emissions reductions beyond those specified in the CAA case
are to occur in the airshed as a result of EPA or state SIP policies. In this case, the Bay
jurisdictions might wish to decide what further NOx emissions reductions from only the Bay
jurisdictions are in their best interests. The Bay community might focus on cleaning up the
Bay, seeking NOx emissions reductions within the Bay jurisdictions to meet nitrate loading
reduction goals at least cost. Or the community might also want to count possible ancillary
benefits from ozone reductions. Where should their efforts at cleaning up air sources be
placed? Within what type of policy? On which sources?
As noted above, there are a number of different approaches to reduce nitrate loadings
to the Bay. These are compared for equivalent nitrate loading reductions -- those registered
by the OTAG scenario, but in this series of analyses applied to the Bay jurisdictions only.
With the controls mandated by that scenario, nitrate loadings to the Bay are forecast to fall by
7 million pounds a year. As in the airshed analysis above, we compare this scenario to the
three other scenarios. The analysis minimizes the costs of reducing nitrate loadings across the
Bay jurisdictions and DC.
Table 16 compares the least cost approach and the emissions trading scenario to the
OTAG scenario. We find that the choice of the least-cost and emissions trading scenario doesn't
really matter, primarily because the nitrate source-receptor coefficients are not much different
among the Bay jurisdictions. Savings are over 50 percent between either of these scenarios and
the OTAG scenario. In particular, the least cost allocation of reductions would attain this goal
at only 44 percent of the OTAG allocation, while resulting in NOx emissions reductions that are
almost identical to OTAG) (310,000 tons for OTAG and 303,000 tons for the least-cost
scenario). Note also that the marginal cost of obtaining the last unit of nitrate loading
reductions to meet the nitrate goal is, at $91/pound, about three times the average cost of $32.
47
Table 16. Summary Results from OTAG and Least Cost Scenarios for the Bay States
Total Nitrate Reduction = 7 million lbs/yr for all Scenarios
Regions
Sources
Total
MD
VA
PA
Mobile Utilities
and
Sources
DC
OTAG
Share of Nitrate Loading Reduction
(% of total)
NOx Emissions reductions
(thousands of tons)
Cost ($millions)
Cost/pound nitrate loadings reduced
I. Least Cost for Nitrate Goal
(minimize gross costs)
Share of Nitrate Loading Reduction
(% of total)
NOx Emissions Reductions
(thousands of tons)
Cost ($millions)
Cost/pound nitrate loadings reduced
Net cost/pound reduced
(high benefits)
Shadow Price ($/lb Nitrates)
II. Least Cost for Nitrate Goal
(minimize net costs, high benefits)
Share of Nitrate Loading Reduction
(% of total)
NOx Emissions Reductions
(thousands of tons)
Gross Cost ($millions)
Ozone Benefits ($millions)
Gross cost/pound nitrate loadings reduced
Net cost/pound nitrate loadings reduced
(high benefits)
Shadow Price ($/lb nitrates)
III. Emissions Trading
(minimize net costs, high benefits)
Share of Nitrate Loadings Reduction
(% of total)
NOx Emissions reductions
(thousands of tons)
Cost ($millions)
Gross cost/pound nitrate loadings reduced
(High Benefits)
Net cost/pound nitrate loadings reduced
(High Benefits)
Shadow Price ($/ton emissions)
Nonutility
Point
Sources
100%
32%
18%
50%
2%
85%
13%
310
67
67
176
39
264
7
$517
$73
$96
$42
$115
$88
$306
$87
$96
$102
$356
$60
65
$434
100%
34%
20%
45%
20%
74%
6%
303
71
72
160
59
223
21
$227
$32
-$4
$36
$15
$38
$26
$153
$48
- $26
- $19
$229
$44
$24
$48
100%
33%
18%
48%
23%
71%
6%
307
70
66
172
71
216
20
$245
$332
$35
- $12
$32
$64
$14
- $14
$36
$68
$27
-$25
$177
$200
$52
-$5
$4
$80
$3
-$48
$219
$233
$44
-$3
$22
$19
$47
$6
100%
27%
22%
50%
26%
68%
6%
317
58
78
180
84
213
19
$264
$37
$11
$6
$60
$38
$193
$55
$32
$18
$210
$44
$21
$47
-$12
-$22
-$14
-$5
-$37
-$4
$7
$91
$38
$781
48
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
The state-by-state allocations of NOx emissions reductions across these scenarios are
not much different, either. Maryland and Virginia reduce NOx more and Pennsylvania
reduces NOx less in the least-cost scenarios relative to the OTAG scenario. Rather, the
principal differences between the least-cost and OTAG scenarios are the greater reliance on
mobile source controls and non-utility point source controls (with correspondingly less
reliance on utility controls) in the least cost scenario, for each of the Bay jurisdictions.
For the Bay jurisdictions the mobile source share of nitrate loading reductions goes
from just 2 percent in the OTAG scenario to 20 percent in the (gross) least cost scenario.
These increases are mirrored by a fall in the utility share from 85 percent to 74 percent. The
shift away from utilities toward mobile sources occurs because, as we noted above, there are
mobile source control options not mandated in the OTAG scenario that deliver NOx
emissions reductions at negative costs relative to the mandated option. Costs per pound of
nitrate loadings reduced are therefore negative for mobile sources (-$19 per pound); they are
about equal ($44 vs. $48 per pound) for utilities and non-utility point sources.
Table 16 also shows what happens when high ozone benefits are included in the leastcost decision on allocation of NOx emissions reductions. Net costs are negative: $87 million
in "net" benefits, with abatement costs of $245 million and health benefits of $332 million.
Compared to the "gross" least-cost scenario where ozone benefits are not included in the
allocation decision, costs are a bit higher (the latter scenario showed costs of $227 million),
although the total emission reductions are not much different (303,000 tons in the gross cost
scenario and 307,000 in the net cost scenario). Note that the shadow price is still positive
(i.e., the costs to society of meeting the nitrate reduction goal are positive at the margin at
$38/pound), but negative on average (-$12).
What shifts in NOx abatement does concern for ozone benefits bring about? Virginia's
NOx emissions reductions are smaller and Pennsylvania's are larger when there is the additional
concern about ozone factored into the allocation decision (the net-least cost scenario). This is
because of the greater ozone benefits that Pennsylvania's NOx emission reductions can bring
compared to Virginia. There is less reliance on utility NOx emissions reductions in the "net"
least-cost scenario (the utility share falls from 85 percent to 70 percent of the total reductions),
with more reliance on mobile sources. Note that benefits exceed costs for each of the states and
for mobile and utility source types but not for non-utility point sources.
Table 16 also shows that the emissions trading scenario, in the aggregate and with
"high" benefits, is more or less equivalent in this instance to the "net" least-cost scenario (with
high benefits), in the sense that average gross and net costs per nitrate loadings reduced are
nearly identical. This result suggests that a simple one-for-one NOx trading system does
relatively well at meeting nitrate loading goals cost-effectively. However, the allocation of
NOx emissions reductions across states is somewhat different. Maryland's NOx emissions
reductions are 83 percent of what they were in the least-cost scenario, while Virginia's are
18 percent larger. Across sources, mobile sources control much more with emissions trading
than in the least-cost scenario. Thus, moving to an emissions trading system from a least-cost
49
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
system results in some state and source reallocations. But these are minor compared to the
differences between these two scenarios and the OTAG scenario.
Policy Analysis 3: National Policy and the Bay Community's Stake
The Bay jurisdictions have a stake in the outcome of the national debate over NOx
controls to meet the ozone standards in the eastern U.S. This debate has been fed by the
OTAG process, in which a wide variety of stakeholders in the eastern U.S. met over a two year period to develop models and policy insights addressing the long-range ozone transport
issue. The debate became more focused with EPA's recent issuance of a proposed SIP call,
requiring specified NOx emissions reductions of each state, and of a NOx trading guidance
document detailing EPA's plan for obtaining NOx emissions reductions cost-effectively
(EPA, 1997; EPA, 1998).
The results presented in this section are designed to help the states affected by these
EPA actions, particularly the Chesapeake Bay jurisdictions, to formulate their responses to
EPA's NOx emissions reduction strategies. Accordingly, we have recast the problem
examined above, that of minimizing NOx abatement costs (with or without netting out
ancillary ozone-related health benefits) to meet nitrate loading reduction goals, to one in
which NOx abatement costs are minimized for meeting either aggregate NOx emissions
reduction targets -- in effect, the EPA emissions trading plan -- or ozone exposure reduction
goals.27 To address the Bay jurisdictions' particular concerns, we note the abatement costs
and ozone benefits of alternative policy approaches as they affect the Bay jurisdictions as well
as noting the nitrate loadings reductions that are an ancillary consequence of a NOx emissions
reduction strategy to reduce ozone. It is important to note that our database is not identical to
that which EPA used to undergird its SIP call and Trading Guidance Documents.28 Thus,
quantitative comparisons between EPA's SIP call and Guidance with our results are not
advised. However, our results provide an indication of the possible consequences of this
program relative to a no trading program and to a variety of variations on the trading theme.
27 More specifically, the emissions constraint problem is defined as finding the allocation of ozone season NOx
emissions that minimizes the annual costs of reducing these emissions subject to the constraint. This approach
leads to a higher cost per ton removed than would be the case if we counted NOx emissions all year round or
adjusted the cost estimates for each technology to seasonal costs. The Pechan database does not permit the latter
to be done in a fully defensible way. The emissions constraint is computed from the Pechan database, which
provides estimates of NOx emissions per day of the ozone season for all sources and their NOx control
technologies. Nitrate loading reductions are tallied assuming that NOx emissions are reduced all year long.
Nitrate reductions for all scenarios would be reduced by a scalar if we assumed that the NOx emissions controls
were shut off in the non-ozone season.
The ozone exposure constraint problem is defined as finding the allocation of ozone season NOx emissions
that minimizes the annual costs of reducing seasonal NOx to meet an ozone exposure constraint. This constraint
is defined in terms of the sum of the products of reductions in ozone concentrations per person per day at each
receptor, multiplied by the population exposed in each receptor and 153 days in the ozone season.
28 Particular examples include: our domain is only 13 states plus D.C., EPA's is 22; the emissions and cost
databases are different.
50
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
The NOx control policies we consider have basically four "margins," or design
parameters: policy type (OTAG-CAC, one-for-one emissions trading, and ozone exposure
trading); participants (some states may decide to form a trading zone with no trading across
the zone); source type (EPA recommends only utilities in the program, but other options
would add other point sources and even mobile sources); and the underlying S-R matrix used
to measure the spatially differentiated effects of NOx on ozone.
There are three groups of results. The first examines variations on EPA's proposed
NOx trading plan, such as the source types participating; the second examines the effects of
using "trading ratios" to guide emissions trading between sources; and the third considers a
"two-zone" option, in which the Bay jurisdictions "go it alone," trading amongst themselves
but not with non-Bay jurisdictions (and vice-versa). We consider least-cost and emissions
trading options in each group.
1. The EPA's Proposed Trading Plan
EPA's SIP call set emission reductions by utility, point, and mobile sources beyond
CAA baselines, corresponding to our OTAG scenario. The agency then issued proposed
guidance for how a NOx trading plan could be constructed that would help some sources meet
these goals at reasonable costs. The "key" features of the plan (key in terms of the modeling
capabilities we have) are that only utilities are in the trading program and emissions trade onefor-one between utilities, even though a ton of emissions in one area may do more or less
damage than a ton of emissions in another area. This simplification mirrored that made for
SO2 allowance trading in Title IV of the CAAA and was justified by EPA on grounds that
introducing trading ratios other than one-to-one would make the trading system too complex.29
2. EPA's Plan and the OTAG Scenario
The most basic comparison is between the OTAG scenario and EPA's plan, given in
columns (2) and (3) on Table 17. This comparison is made by using the utility NOx
emissions from the OTAG scenario as the constraint for the cost minimization model. For
EPA's plan, non-utility point and mobile sources are assumed to perform according to the
OTAG scenario. The total cost of the OTAG scenario is $1.726 billion, as noted in VI.A
above, where utilities shoulder $1.1 billion of this cost. With EPA's plan, in which the utility
reductions can be obtained by inter-utility trading of emissions anywhere in the domain,
utility costs fall about 13 percent to $969 million. Average cost- effectiveness, in terms of
NOx emissions reductions by utilities falls from $2,788 to $2,425 per ton. That they don't fall
more is probably because the performance standard is very tight, so there is little opportunity
for gains from trade.
29 EPA's "SIP call" actually requires utilities to meet a 0.15lb NOx/mmbtu standard and 70 percent reduction for
large non-utility boilers, whichever is less. Our modeling assumes, in contrast to EPA, but more in line with the
OTAG process, that utilities may meet the less stringent of the 0.15lb/mmbtu standard or 85 percent reduction
and 70 percent reduction for the large non-utility boilers.
51
Table 17. Consequences of Alternative NOx trading plans to Meet Aggregate NOx Emissions Reduction Targets
OTAG
Airshed Abatement
Costs ($ millions)
Airshed Utility Costs
($ millions)
Airshed Ozone Exposure
Reductions
Bay jurisdictions
Abatement Costs ($
millions)
Bay jurisdictions (MDVA) Ozone Benefits
($millions)
Nitrate Loading
Reductions (millions of
lbs.)
EPA Plan:
ET Adding ET Adding
Emissons
Other Point Mobile
Trading (ET) Sources26
Sources
with Utilities
only1
1,726
1,581
1,245
987
1,114
969
888
728
41,612
41,580
41,596
517
472
22.9
13.7
Ozone
Exposure
Trading (OT)
with Utilities
only: Average
Episode
1,578
OT with
Utilities only:
NE Episode
OT with all
sources
(6a)
1,379
805
966
767
436
42,060
41,580
41,580
42,060
412
327
473
403
246
21.5
23.0
22.4
19.5
16.6
24.2
12.7
13.3
13.9
11.6
11.6
11.0
1 Sources excluded in this trading scenario are assumed to be under the OTAG plan.
52
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
Should the Bay jurisdictions support EPA's plan over the OTAG scenario? Abatement
costs are lower, with all of this cost savings accrued by utilities. So, on a cost basis the plan
deserves support. Yet, nitrate loadings to the Bay fall more with the command and control
approach -- 13.7 million pounds a year versus 12.7 million pounds -- combined with higher
ozone benefits than for EPA's plan. Thus, the Bay jurisdictions face a tradeoff between costs
and nitrate loading reductions when choosing between these two options.
3. Variations on EPA's Plan
Would variations on EPA's Plan be better for the airshed and the Bay jurisdictions?
We examine three variations: a plan that adds other point sources to the trading program, a
plan that adds mobile sources to the previous plan, and a plan that introduces emissions
trading at ratios dictated by runs of the Urban Airshed Model-V, which we term ozone
exposure trading (OT).
Adding Point Sources. Allowing additional source-types to trade should reduce
overall abatement costs of meeting the aggregate emission reductions target, but the effect on
the Bay jurisdictions is not predictable, a priori. Column (3) shows that adding other point
sources further lowers aggregate abatement costs from $1.581 billion with utility trading
alone to $1.245 billion, lowering utility costs from $969 million to $888 million. From the
Bay jurisdictions' perspective, this program has lower abatement costs and larger nitrate
loading reductions, so it clearly dominates the utility-only trading plan.
Adding Mobile Sources. It will obviously be technologically difficult, if not
impossible, to incorporate all mobile sources into a trading program in the foreseeable future.
However, some states already have programs involving mobile source and point source
trades, called Mobile Emission Reduction Credit Programs, and one can imagine trading
programs involving fleet vehicles. In any event, modeling this type of program provides an
estimate of potential benefits from incorporating this source type into trading.
The result is a rather dramatic reduction in airshed-wide abatement costs compared to
the utility plus other point source trading program, with costs falling from $1.245 billion to
$987 million and further cost savings to the utilities. Costs for the Bay jurisdictions fall
dramatically as well when mobile sources are added because other regions foot the bill. A
surprising finding is that nitrate loading reductions are actually greater with this plan (13.9
million pounds) than with any of the above approaches, including the OTAG scenario. Ozone
benefits are about the same as with the other approaches. The combination of much lower
abatement costs and greater nitrate loadings reductions makes this all-source trading scenario
a possible winner.
Departing From 1:1 Trades. Section III made it clear that reductions in NOx
emissions are more productive in reducing ozone exposures from some locations than from
others, and that this "productivity" varies by episode type. We model several variants of a
trading program to capture such effects. The first (column 6) returns to the utility-only
program, where the trading ratios are taken from the source-receptor coefficients applicable to
53
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
the average of the NE, MW, and SE episodes and the constraint is the ozone exposure
reductions obtained from reducing emissions under EPA's planned program (column 2).
Surprisingly, the utility-only, average episode trading policy performs no better from
the airshed perspective, and worse over our three measures from the Bay jurisdictions'
perspective than EPA's one-for-one NOx trading plan. However, this result is not robust to
the episode chosen. If, instead, trading ratios are defined by the northeast episode (column 7),
then utility abatement costs to meet the identical standard ($767 million) are quite a bit lower
then the EPA Plan ($969 million), with Bay jurisdiction costs falling as well. Costeffectiveness in terms of NOx emissions reductions is now only $2,153.
The second variant is an all-source ozone exposure trading program. This program
(column 7), in comparison to the all-source emissions trading program (column 5) saves
another $180 million and reduces utility costs to $436 million. The Bay jurisdictions face a
tradeoff between lower costs and higher ozone benefits on the one hand and lower nitrate
loadings reductions on the other.
4. Sensitivity to alternative ozone episodes to meet ozone exposure reduction targets
A number of questions are raised by the above results showing how the consequences
of using the NE or the average episode to define trading ratios are so different. On the one
hand, the advantages of "spatially differentiated" permit trading systems are well known.
Conceptual work by Tietenberg (1985) and others proves the efficiency benefits of designing
tradable permit systems that take into account the differential location effects of emissions on
concentrations. On the other hand, whether such a more complex system is "worth" the
trouble is strictly an empirical issue. In addition, this literature is silent on the issue of how to
establish such ratios in the presence of modeling uncertainty and meteorological variability.
Clearly, the first step is to establish whether choosing alternative S-R coefficients for the
trading ratios leads to wildly different results.
Table 18 shows the consequences of meeting ozone exposure targets by trading
emissions at ratios dictated by the UAM-V for various meteorological conditions for both a
utility-only and an all-source permit system. Columns (2) and (3) show the NE and average
episode results again, but add results for an all-source program. Results using the other
episode types are in columns (4) and (5).
Examining the utility-only program first, we see the low costs associated with the NE
S-R coefficients, an indication of how productive NE meteorological conditions are for
reducing ozone. The MW episode and the average episode are the most similar in their
consequences. And defining trading ratios according to the SE episode results in the target
not being reached, even with all utilities maximally reducing their NOx emissions.
From the Bay's perspective, we see the usual tradeoffs. Nitrate loading reductions are
larger for the most costly scenario to the airshed (11.6 million pounds for the average episode
compared to 10.4 million pounds for the NE episode). From a cost perspective, however, the
Bay jurisdictions want the most efficient approach for the airshed, which is to trade according to
the NE episode. Ozone benefits to the Bay are largest with the MW episode. Comparing these
54
Table 18. Consequences of Alternative Trading Ratios for Meeting An Ozone Exposure Reduction Target
(from CAC). Utilities only (all sources)
Average
Episode
Airshed Abatement
Costs ($millions)
Bay jurisdiction
Abatement Costs
($millions)
Bay jurisdiction (MDVA) Ozone Benefits
($millions)
Nitrate loading
reductions (millions of
lbs.)
Airshed Abatement
Costs ($millions)
Bay jurisdiction
Abatement Costs
($millions)
Bay jurisdiction (MDVA) Ozone Benefits
($millions)
Nitrate loading
reductions (millions of
lbs.)
NE Episode
MW
Episode
SE Episode
CAC
Average
Episode
Infeasible
1114
966
Utilities-only
767
916
310
241
311
I
356
19.5
16.6
26.2
I
24.8
11.6
10.4
11.4
I
11.6
878
1726
789
All Source-Types
617
843
240
230
287
181
520
24.2
26.7
30.8
10.6
27.5
10.9
10.0
11.8
11.1
13.7
55
CAC
NE
Episode
27.2
31.6
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
trading options to the corresponding OTAG scenario, we find lower costs and lower ozone
benefits to the Bay jurisdictions with the trading options compared to the OTAG scenario.
Thus, the standard trade-offs are much in evidence, with nitrate loadings not much of an issue,
except for the relatively low loadings reductions for the NE episode.
Because mobile sources have different S-R coefficients than point sources, it would be
no surprise to find the story somewhat different when all sources are considered. Indeed,
while the NE episode is still the cheapest, this is followed by the average, then the MW, and
then the SE episode, the last being feasible when all source types are in the program. The
implications for the Bay are different as well. The most surprising finding is that even though
trading according to ratios based on the SE episode is the most expensive for the airshed, it's
the cheapest for the Bay jurisdictions, who do not participate very much given their low
effects on ozone throughout the domain under the SE conditions.30 Furthermore, nitrate
loading reductions are actually larger with trading ratios from the MW episode than from the
SE episode. Note how with the SE episode, ozone benefits within the Bay jurisdictions are
very small. So, the Bay jurisdictions face particularly interesting tradeoffs in supporting one
set of trading ratios over another.
What implications does this sensitivity of results to trading ratios have for tradable
program design from the airshed perspective? If attainment is what matters (and we are NOT
modeling attainment here), then, given that the NE episode is associated with most
nonattainment (at least for the current standard), the trading ratios should be set for a NE
episode to deliver the most ozone reductions when and where they are most needed. At the
same time, if we are most concerned about maximizing the ozone reduction benefits per ton
of NOx reduced (which is, of course, what we are ultimately concerned about), then with
NOx emissions reductions being the most "productive" during NE conditions, the NE trading
ratios should be adopted. Thus, in both situations, the NE ratios are the most promising. This
is a lucky result, meaning that the program can be designed to take advantage of a win-win
situation.
There is a downside to dealing with this meteorological variability and cost sensitivity
by using the NE episode -- during a SE or MW episode, the exposure target probably won't be
met. To illustrate the problem, we found the least-cost allocation of NOx emissions to meet
the ozone exposure constraint using the NE episode S-R coefficients and then calculated
ozone exposure reductions given the NOx emissions reductions allocation. These runs of the
model show (Table 19) that the occurrence of a MW episode and a SE episode, respectively,
when trading ratios are based on the NE episode, results in ozone exposure targets being
missed in the airshed by 20 percent and 22 percent, respectively. The estimates for the Bay
are also included. The MW episode can clearly cause the greatest losses to the Bay, if trading
ratios are established on the basis of the NE episode.
30 Note that trading ratios are administratively established and drive the NOx allocations. However, ozone
benefits are related to actual meteorological conditions that will differ from any administratively determined
ratios.
56
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
Table 19. Loss in Benefits from a Mismatch between Trading Ratios and Episode Experienced.
Episode Type
NE
SE
MW
AVE
Domain
Ozone Exposures
(billion person*
ppb*day/ozone
season)
49.1
38.4
39.2
42.3
Bay
% of NE
78
80
86
Ozone Exposures
(billion person*
ppb*day/ozone
season)
14.9
17.4
8.0
13.4
% of NE
85
54
90
5. Two Zone Trading
Table 20 presents the results of dividing the domain into two trading zones--one with
the Bay jurisdictions, the other with the other states. Such a program might be implemented if
the Bay jurisdictions, in their concern about nitrate loadings, were uneasy about permitting
NOx trades from sellers outside the area to buyers in it, although supportive of programs
where the Bay jurisdictions are net sellers of NOx. While the Bay jurisdictions might be glad
to "export" pollution to the non-Bay jurisdictions to help reduce nitrate loadings, the non-Bay
jurisdictions are not likely to permit this one-way arrangement. So if the Bay jurisdictions
pressed their plan, the only stable outcome is where the Bay jurisdictions meet their ozone
reduction or emissions reduction targets alone, creating, in effect, two trading zones with no
trade between them.
Table 20. Single zone versus two-zone trading to Meet Ozone Exposure Targets.
All source types
2
2a
Domain-Wide Trading
Bay
Non-Bay
Total
Abatement Costs
($millions/year)
Emission
reductions
(tons NOx/year)
Nitrate Loading
Reductions
(lbs Nitrate/year)
2b
Two Trading Zones
Bay
Non-Bay
Total
243.9
545.5
789.4
248.7
540.7
789.4
120.1
268.4
388.5
121.2
267.3
388.5
5,529
5,369
10,898
5,668
5,369
11,037
57
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
Conceptually, a two-zone program can be expected to be more expensive, in the
aggregate, than a one-zone program, but a comparison of other effects is ambiguous
conceptually. For instance, if sources in the Bay "zone" have lower marginal costs than those
outside this zone, then, in effect, there will be NOx emissions on net being traded from within
to outside of the Bay zone. If sources within the Bay zone have higher marginal costs than
those outside this zone, the trading will be in the reverse direction on net.
The results (in table 20) show that aggregate abatement costs are virtually identical in
the one zone and two zone situations, indicating flat marginal cost curves around the implied
control levels. However, in the two zone situation, the Bay's sources shoulder more of the
abatement costs than in the one-zone situation and the sources in the non-Bay zone shoulder
correspondingly less. Nitrate loading reductions are higher in the two zone situation because
emission reductions are greater for sources in this zone. The same issue can be investigated
for an emissions trading system, in this case EPA's plan, i.e., including utilities only, for one
zone and for two zones. The results are identical qualitatively. So, the Bay jurisdictions face
the tradeoff of higher abatement costs for greater nitrate loading reductions if they elect to
form a separate trading zone.
VII. CONCLUSIONS AND POLICY IMPLICATIONS
Airborne nitrates contribute from 20 to 30 percent of the nitrate loading to the
Chesapeake Bay. Given the interest in policies for reducing NOx emissions to meet ambient
air quality standards for ozone (and particulates), it is useful to examine how the parallel
concern over nitrate deposition to the Bay affects the appropriate allocation of NOx emissions
reductions among utilities, non-utility point sources, and mobile sources.
Our analyses of the issue are of three types. The first is a detailed analysis of the
effect on nitrate loadings to the Bay of command and control (CAC) policies specified in the
CAA and as part of the OTAG process. The second is a comparison of alternative scenarios
for reducing NOx emissions that meet nitrate loading goals, with or without concern for
reducing ozone concentrations and the health effects they cause. The third is a comparison of
alternative approaches to allocate NOx emissions to meet ozone exposure goals while tallying
the ancillary effect on nitrate loadings. It therefore focuses on the stake that the Bay
jurisdictions have in the outcome of negotiations over NOx trading programs being developed
by EPA for reducing ozone in the Eastern U.S.
This analysis is unique in several respects -- in its exploration of quantitative estimates
of the lower cost alternatives to current regulatory programs, in its integration of ancillary
benefits, in its focus on the Bay jurisdictions, and in its examination of these ancillary benefits
for alternative meteorological episodes.
The most basic message to the Bay community is that the Chesapeake Bay obtains a
bonus from efforts to reduce NOx emissions to meet the ambient air quality standard for
ozone. In our analysis, airborne NOx emission reductions slated to occur under the Clean Air
Act by 2005 in the 13 state (plus D.C.) domain will reduce nitrate loadings to the Bay by
about 27 percent of the baseline airborne loadings of 66 million pounds per year. The
58
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
additional control of NOx contemplated in our OTAG scenario (0.15lb NOx/mmBTU for
utilities or 85 percent removal, whichever is less and 70 percent reductions for large point
sources; LEVs for mobile sources) is estimated to result in an additional 20 percent reduction
from this baseline. This bonus is not a panacea, however, because final decisions on the
additional levels of NOx control have yet to be made, and, even if these estimates are
accurate, the approximately 47 percent reduction in associated nitrate loadings constitutes
only about a one-sixth reduction in the controllable nitrate loads from all sources in the Bay.
Clearly, the costs of obtaining such reductions can be significantly reduced by
rearranging the allocation of emissions reductions to take advantage of source-type and
locational considerations. This paper has not focused on the policies needed to obtain such
savings but has pointed to the locations and technologies that have the greatest potential to
achieve the targets in lower cost ways. In addition, we find that adding consideration of
ancillary ozone-related health benefits to the picture does not alter any qualitative
conclusions. Quantitatively, unless a link between ozone and mortality risk is assumed, the
benefits are too small to affect the cost-saving allocations of NOx reductions. With such a
link, NOx reductions are larger in total and shift to non-Bay states and towards utility sources.
These specific effects are sensitive to the source-receptor coefficients linking NOx to ozone,
however.
Our analyses also suggest that the Bay jurisdictions have a stake in the outcome of the
NOx trading debate -- that some trading designs can lead to better outcomes for these
jurisdictions than others. Nevertheless, a common feature of cost-savings policies is that they
both rearrange emissions reductions and, in the aggregate, reduce emissions less than a
command and control system. Thus, some trading regimes result in significantly smaller
loadings reductions (up to 25 percent smaller) than the command and control approach.
A. Regulatory Analysis (CAA and OTAG Scenarios)
Both in implementing CAA mandates and in going beyond such direct mandates to
implementing controls suggested by our OTAG scenario, the Bay jurisdictions (Pennsylvania,
Maryland, Virginia, and the District of Columbia) are a relatively inexpensive source of
nitrate loading reductions, measured in $/nitrate loading terms. This result is due almost
entirely to the locational advantage of these states, i.e., their NOx emissions reductions
produce greater reductions in nitrate loadings to the Bay than NOx emissions reductions from
more outlying states. However, in some cases, NOx emissions reductions from sources in the
Bay jurisdictions are less cost-effective than reductions in other states when effectiveness is
measured as $/ton of emissions. Utilities are responsible for the vast majority of the NOx and
nitrate loading reductions.
An important extension of the analysis of current regulatory programs of this paper is
to take into consideration both the air and water impacts of controls in the cost-effectiveness
analyses. Clearly, not all of the costs of NOx emissions reductions that are undertaken
primarily to improve air quality should be attributed to nitrate loading reduction in the Bay.
We develop a method for netting the ancillary benefits from NOx emissions reductions costs
59
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
to develop a more inclusive measure of net social costs of NOx emissions reduction policies.
Are the above conclusions changed if some of these effects -- the ozone-related health
benefits from NOx emissions reductions -- are added to the picture? In general, the
conclusions are not altered when ozone-related health benefits are considered. With our best
estimate of the health benefits per unit ozone exposure reductions, the health benefits are too
small to matter.
However, there is significant controversy in estimating ozone-related health benefits,
particularly concerning the effects of ozone on mortality risks. We have developed a "high"
estimate of ozone benefits per person-ppb exposure change that incorporates effects on
mortality -- a benefit category ignored in our "best" benefit estimates. However, even using
this high estimate, the cost-effectiveness of NOx emissions reductions in the non-Bay
jurisdictions versus the Bay jurisdictions is not much altered. This occurs because the S-R
relationships linking NOx emissions to ozone changes over the ozone airshed are similar
across states. Therefore, states like Kentucky and Maryland have a similar effect on
aggregate ancillary health benefits, which does not alter the cost-effectiveness comparison
based on net rather than gross cost terms. At the same time, each state's cost-effectiveness
drops dramatically when the "high" benefit estimate is netted out of abatement cost. For the
airhsed, the aggregate cost effectiveness falls from $126/lb nitrate to $54/lb nitrate loading
when moving from the CAA to the OTAG scenario.
B. Comparison of OTAG Regulatory Scenario to Alternative Scenarios for Meeting
Nitrate Loading Targets
As shown in table 12, significant cost savings to meet nitrate loading reduction goals
are potentially available by improving the targeting of NOx emissions reduction policies to
take advantage of cost-differentials across sources and locational differences in the creation
and transport of ozone and nitrate loadings to the Bay. The limit of such cost savings (under
the least-cost scenario) is over $1 billion annually (for utility, point source, and mobile
sources) from the OTAG regulatory scenario costs of $1.7 billion. These savings can be
realized by shifting NOx abatement to mobile sources, such as through enhanced I&M,31
relaxing the 0.15 lb NOx/mmBTU requirement on utilities and large point sources (such as
through a trading program) and shifting abatement to sources of all types within the Bay
jurisdictions.
In attempting to meet nitrate loading reduction goals while minimizing net costs
(abatement costs minus ancillary "high" ozone-related health benefits) emission reductions
and abatement costs are larger than those in the least-cost scenario. But, ozone benefits are
far larger than they would be if the allocation of NOx were done to meet nitrate loading goals
alone. At the same time, there is a important regional aspect to the comparison of NOx
31 The sensitivity to this result needs to be explored in future work. There are very large emission reductions
assumed to occur at very low cost under enhanced I/M in the MOBILE Model (see Appendix 3). These
assumptions have recently been called into question (Harrington, McConnell and Cannon, 1998).
60
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
emissions reductions in this scenario to one without ancillary benefits considered. In
particular, eastern Pennsylvania and New Jersey exhibit ozone scavenging in all of the ozone
episodes we considered. This results in a "bias" against NOx emissions reductions for these
states. For the NE (Bermuda high) episode, the bias is towards NOx emissions reductions in
Ohio and New York.
While the least-cost scenario can provide a sense of the limits of cost savings over the
OTAG scenarios, our emissions trading scenario provides, perhaps, a more realistic view of
such savings. In this scenario, in effect, emissions are traded across regions and sources tonfor-ton, but the emissions cap is set (through a series of optimization runs) to meet the same
nitrate loading goal for the Bay that was used for both the OTAG scenario and the least cost
scenario described above. This scenario delivers substantial cost savings over the OTAG
scenario, but costs and emission reductions are larger than they are in the least-cost scenario.
The reductions of NOx shift much more away from the Bay jurisdictions to states in the
Midwest.
Alternatively, if the emissions trading program to meet the nitrate loading goal is
made more complex by accounting for ancillary ozone benefits (high benefits case) at the
same time,32 the nitrate loading goal can be met at much lower cost per ton than can be
obtained from an emissions trading program alone.
Because it may be difficult to envision a trading program involving states outside the
Bay partners in the Chesapeake Bay Agreement, we also examined the consequences of
implementing a variety of policies to meet nitrate loading goals only within these Bay
jurisdictions. The conclusions for this "Bay jurisdiction" analysis are qualitatively the same
as those for the above "airshed" analysis, with the exception that the results of an emissions
trading program look much more like those in the least-cost scenario. This occurs because the
least-cost scenario takes locational differences in the NOx to nitrate loading relationships into
account while the emissions trading scenario does not, and these relationships are very similar
across the Bay jurisdictions but not across all states in the airshed.
Finally, we looked at the impact of different ozone episodes on the least cost
allocations for achieving the nitrate loading reduction goal. The type of episode causes costs,
emission reductions and allocation across sources to vary, but not dramatically. Of the three
types of ozone episodes, the Northeast episode stands out as requiring the most NOx
emissions reductions to meet the target. As a result, it is more costly to reach the nitrate
loading goal under the Northeast episode than under the other episodes (about 10 percent
more expensive). In addition, for many regions controlling utilities is more cost-effective
during the Northeast episode, so if policy were to made based on those events there would be
a shift toward more utility reductions and fewer mobile source reductions.
32 One way to do this is through a tax on NOx emissions that internalizes the ancillary ozone benefits.
61
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
C. The Bay Jurisdictions' Stake in NOx Trading for Reducing Ambient Ozone
This series of scenarios is motivated by EPA's recently issued proposed Guidance
Document for a NOx trading program to help meet ambient ozone standards in the eastern
U.S. It takes the point of view of the Bay jurisdictions, asking whether particular design
features of this program deserve support from the Bay jurisdictions, in terms of the ancillary
effect on nitrate loadings, ozone health benefits enjoyed by people living in the Bay
jurisdictions, and costs to sources within the Bay jurisdictions. The results cannot be
quantitatively compared to EPA's modeling efforts in support of the guidance document, or to
modeling results produced by the OTAG process, however. Our domain is only 13 states plus
D.C., while the EPA domain includes an additional eight states; we do not consider area
sources, while EPA does; the emissions and cost database we used from Pechan is not
identical to the database used by EPA, particularly with respect to electric utilities; and our
ozone episodes are different than EPA's, although we both consider Bermuda High-type
meteorology. Most important, our analysis is designed to meet not only emissions reduction
targets, as EPA's, but also aggregate ozone exposure reduction goals.
In spite of these differences, we feel that our modeling results are qualitatively
comparable to EPA's. For instance, our analysis supports the EPA analysis that a utilitiesonly emissions trading system saves money over the OTAG scenario applied to all sources.
We find, for instance, that overall costs are $1.581 billion with emissions trading relative to
$1.726 billion with our OTAG regulatory scenario, where all of the savings is experienced by
the utilities in the program (whose costs fall from $1.114 billion to $969 million annually).
The overall conclusion from this analysis is that the Bay jurisdictions should not be
indifferent to design features of a NOx trading program. Along some design dimensions, one
option may dominate. More frequently, there is a tradeoff to the Bay jurisdictions between
gaining greater nitrate loading reductions and bearing greater costs, with ozone benefits not
being determinative.
In terms of source participation, the Bay jurisdictions benefit more when the trading
program includes point and mobile sources (in this paper we are not addressing how such a
program could be structured, or even the extent to which it might be possible). This program
delivers more nitrate loading reductions, no lower ozone benefits, and lower costs to Bay
jurisdiction sources than either EPA's plan, a plan that adds point sources, or the OTAG
scenario.
A number of analysts have been urging EPA to modify their proposed trading system
to account for locational and source-specific differences in the productivity of NOx emissions
reductions in creating ozone and in the transport of ozone to other regions. Account could be
taken by altering the trading ratios for emission trades between sources in different regions
and of different types. We examined this issue and found that substantial cost savings are
possible, over and above those available from emissions trading. If locational differences
could be fully taken advantage of, costs would fall to $805 million for an all-source trading
ratio program, relative to $987 for an all-source one-for-one trading program.
62
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
Interestingly, whether EPA's utility-only trading program should depart from a onefor-one design depends on the type of episode underlying this program. For the average of
our "typical" episodes, EPA's one for one trading program performs just as well as a trading
ratio program. With NE trading ratios, however, utilities would save $200 million with a
trading ratio program.
This finding prompted us to test the sensitivity of our results against the four episode
types (an average and a NE, MW, and SE episode type). We find that the episode-type affects
costs, emission reductions and nitrate loadings (indirectly); and that the NE episode features
the most productive meteorology for reducing ozone through NOx emissions reductions. As a
result, using this episode to model least-cost ozone reductions and emissions trading yields
the lowest costs to the airshed and the Bay to meet ozone exposure reduction goals. This is a
particularly important result because the NE episode is the one usually associated with
nonattainment of ozone standards. Therefore, attainment considerations and our results all
point towards specifying trading ratios in terms of a NE episode.
Still, because NE conditions occur only during a part of the summer, it is relevant to
consider what losses may arise if conditions differ. We find, first, that with SE meteorology,
it is not even feasible to reach the given ozone exposure reduction goal. Further, through the
domain, the occurrence of other conditions can result in a 15 percent loss in ozone exposure
reductions if the allocation of NOx emissions reduction is optimized for a NE episode but
some other episode-type occurs. The Bay jurisdictions would experience smaller losses
unless a SE episode dominates the summer.
Finally, we envision the Bay jurisdictions deciding that the NOx trading program
designed solely for ozone reductions may not meet their needs. A particular problem might
be that the seasonal program being advanced by EPA will result in nitrate loading reductions
far short of those from a year-round program. Therefore, we examine the consequences to the
Bay jurisdictions of creating a trading zone, leaving the other airshed states to create their
own trading zone (with no trading of emissions between zones). We find that this autonomy
comes at a price, i.e., the cost of the additional nitrate loading reductions to the Bay
jurisdiction sources are fairly high. The sources pay $5 million more than they otherwise
would to obtain an extra 140,000 pounds of nitrate loading reductions, or $36/pound.
However, this is only a small amount below what they would pay in the CAC scenario -$38/pound.
D. Limitations and Caveats
This analysis has a number of limitations that temper the usefulness and credibility of
its conclusions. Additional research is planned to address some of these concerns
(particularly numbers 3,4,6,7, and 8).
1. The source domain is limited to 13 states plus D.C. rather than the 22 or 36
commonly considered in discussion about limiting NOx emissions in the Eastern
U.S.
63
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
2. The receptor domain encompasses New England as well as the source regions;
however, it is narrower than that considered by EPA and in the OTAG process.
3. The Pechan emissions and cost data are not from the latest Pechan dataset. In
addition, the utility data were judged to be less reliable by EPA than the
information contained in the Integrated Planning Model (IPM), which EPA uses in
its own analyses of NOx trading issues. Our own analysis shows that, at least, the
Pechan data contain more abatement options than the IPM dataset.
4. The Pechan mobile source data appear, in light of research conducted at RFF and
elsewhere, to seriously underestimate the costs and overestimate the effectiveness
of NOx emissions reduction for some mobile source options.
5. The cost data were originally developed in such a way as to leave ambiguous the
incremental costs of choosing one abatement option over another. (See Section II.)
6. The S-R coefficients linking NOx emissions to nitrate loadings are not the latest
version from RADM and have some limitations. First, the source regions are
highly aggregated spatially, often to the state level. Second, there is spotty
coverage of some states, or even no coverage, requiring imputations of S-Rs using
a nearby state's data. Third, there is spotty information on S-Rs by source-type,
with some states provided with separate S-Rs for utility, other major point, and
mobile sources, and other states provided with aggregate coefficients. Finally,
there are several internal inconsistencies (see Appendix A: Regional Acid
Deposition Model Summary Output, in Pechan, 1996) in the S-Rs. In addition, the
chemistry of the RADM model has continued to evolve since the underlying
model runs were performed, calling into question the credibility of the S-Rs being
used to date.
7. The ozone S-R coefficients have not yet been subjected to sensitivity analysis or
other tests of their robustness. Such coefficients, even for the same meteorology,
may be sensitive to the starting initial concentrations, the magnitude of the
emissions change, and whether NOx emissions change at more than one region at
a time.
8. Our modeling so far omits consideration of ancillary benefits to NOx emissions
reductions to the extent such reductions lower airborne nitrates, which register as
lower fine particle concentrations. EPA has recently recognized the health threat
posed by such particles in setting a new fine particulate ambient air quality
standards.
9. Ancillary benefit estimates are limited to health. Benefits may also be registered
for materials, for instance. See Lee et al. (1995) for a full discussion of the benefit
pathways for ozone and PM.
64
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
Appendix 1: THE CONVEX HULL AND OPTIMIZATION
This appendix describes the algorithm and mathematical approach used to find the
least-cost allocation of NOx emissions reductions to meet the given constraints. This
algorithm included determination of the "convex hull," which is basically a marginal cost
function containing only technologies not dominated by other technologies, in terms of costs
and emission reductions. It also included using the convex hulls for each source in an
optimization process, which was written in GAMS (General Algebraic Modeling System).
Determining convex hull
Scenarios examining the costs of reduction of pollutants used the 2005 baseline
technologies for each source as the starting point in the determination of the convex hulls.
Similarly, scenarios examining the costs of pollution reduction, given the fact that CAA
technologies were in place, used the 2005 CAA technologies as the starting point in the
determination of the convex hulls. The appropriate cost and NOx emissions reduction of
every possible technology for each source was computed relative to the appropriate starting
point for the analysis. All technologies with NOx emissions reductions less than the NOx
emissions reductions of the starting point technology were dropped from the set of feasible
options. Technologies having greater NOx emissions reductions than the starting point
technology but providing them at a lower cost, and thus having negative marginal costs, were
included in the feasible set. These technologies were included to account for the possibility
that the command and control scenario specified technologies that were less efficient than
other available options. For the remaining technologies, the marginal cost of moving from
one technology to another (the change in cost/change in NOx emissions reductions) of all
technologies relative to the starting point were calculated. The technology having the
minimum slope was selected as the next point on the convex hull after the starting point.
Information for this point (marginal cost, marginal NOx emissions reduction, load to emission
ratio, technology identifiers, etc.) was retained and this point was used as the new reference
point in determining the next point on the convex hull. Technologies having fewer NOx
emissions reductions than the new convex hull point are dominated by other technologies and
therefore, were dropped. The minimum slope from this point was determined, identifying the
next point on the convex hull. The process was repeated until the technology options were
exhausted for each source.
For a given source, cost and NOx emissions reduction information for the technologies
on the convex hull was retained in the form of marginal costs and marginal NOx emissions
reductions. Total cost and NOx emissions reduction can be derived from the marginal values
since:
Total cost of technology i =
65
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
∑ [MCi * ∆NRi ]
i
where ∆NR is the marginal NOx emissions reduction.
Total NOx emissions reduction from technology i =
∑ MCi * ∆NRi
i
∑ ∆NRi
i
When ancillary ozone benefits were incorporated into the analysis, both gross and net (gross
minus ancillary benefits) marginal costs were calculated. The convex hull based on minimum
gross marginal costs (retaining the net MC information) was determined for use in cost
minimization scenarios optimizing over gross costs, while the convex hull based on minimum
net MC (retaining the gross MC information) was determined for cost minimization scenarios
optimizing over net costs.
Optimization Process:
All optimization scenarios minimized the total cost of NOx emissions reduction:
min ∑ c kj x kj
k, j
Where:
k= index of sources
j= index of abatement options on convex hull for a source
c=marginal cost of technology j given technology j-1 for source k
x=marginal NOx emissions reduction of technology j given technology j-1 for source k
The NOx emissions reduction for a given source and technology option was constrained not
to exceed the amount determined by the convex hull:
x kj ≤ d kj ∀k , j
Where:
d= marginal NOx emissions reduction as determined by convex hull
66
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
Scenarios that explicitly accounted for a source's impact on nitrate loadings or ozone included
the constraint:

∑ a  ∑ x
k
k

j
kj

≥ L


Where:
L= nitrate load reduction or ozone reduction from command and control scenarios
a= coefficient converting NOx emissions into nitrate loading or ozone
Scenarios not accounting for sources' impacts on nitrate loadings or ozone used the constraint:
∑x
kj
≥E
k, j
Where:
E=NOx emissions reduction constraint
Using the optimal xkj derived from the constrained cost minimization problem and the
coefficients ak , the corresponding nitrate loading or ozone changes were calculated. If the
calculated load or ozone reduction was less that L, E was increased and the problem was
resolved. This process was repeated until the minimum value of E resulting in a nitrate
loading or ozone reduction as least as great as L. was found.
67
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
Appendix 2: UTILITY COST FUNCTIONS
The Pechan database models a wide variety of NOx control strategies for each point source
boiler type. This is illustrated below for dry-bottom wall-fired coal boilers. Most of the
capital cost equations are variations of the standard form aXb, where X is the maximum
nameplate capacity of the boiler. The operating and maintenance (O&M) cost equations can
have fixed and/or variable cost components. The variable cost component is related to the
maximum nameplate capacity of the boiler, and may or may not be related to the boiler
capacity factor.
Dry-Bottom Wall-Fired Coal Boilers
NOx Controls
LNB (low-NOx
burner)
SNCR (selective
non-catalytic
reduction)
NGR (natural gas
reburn)
LNB + OFA
(overfire air)
LNB + SNCR
SCR (selective
catalytic reduction)
SNCR+SCR
LNB+OFA+SCR
Average
Reduction
(%)
Average
Cost per
Ton
($/ton)
450 MW:
Capital
Cost
($1000/
year)
4,698
300 MW:
O&M Cost
($1000/
year)
450 MW:
O&M Cost
($1000/
year)
180
300 MW:
Capital
Cost
($1000/
year)
4,140
46
127
167
48
975
2,889
3,684
1,159
1,739
50
521
6,355
8,106
1,390
2,085
53
300
5,477
6,215
147
190
65
75
493
1,815
10,870
19,066
13,982
24,318
1,255
2,178
1,772
3,267
88
90
1,315
1,590
13,404
41,066
17,096
53,928
2,804
5,546
4,206
8,183
Note: A capacity factor of 0.6148 was assumed.
68
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
Appendix 3: MOBILE SOURCE DATA
Pechan and Associates, Inc. provided the Mobile Source data for the analysis in this paper.
This dataset included emission reductions and costs for the following control options
(described in more detail in Pechan's Emission Reduction and Cost Analysis Model for NOx
(ERCAM-NOx, 1997):
• One of three possible I/M programs
Basic I/M
Low Enhanced I/M
High Enhanced I/M
• Reformulated gasoline
• Reformulated diesel
• LEV program – the emission reduction from LEVs depends on what type of I/M
program is in place (LEVs get substantially more reductions if they are initiated in
combination with Enhanced I/M; see below for more detail).
There are 7 categories of vehicles listed in Table A3-1 below. The table shows which control
options can be applied to which vehicle types in our analysis.
Table A3-1. Vehicle Type and Control Option Combinations
Heavy
Heavy
Duty Diesel Duty Gas
Vehicles
Vehicles
(HDDV)
(HDGV)
I&M
Reformulated
Gasoline
Reformulated
Diesel
LEV
Light
Duty
Diesel
Vehicles
(LDDV)
Light
Duty
Diesel
Trucks
(LDDT)
Yes
Yes
Yes
Light Duty
Gas
Trucks 1
(LDGT1)
Light Duty
Gas
Trucks 2
(LDGT2)
Light Duty
Gas
Vehicles
(LDGV)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Scenarios and Spatial Aggregation:
Each county in the airshed has some baseline option/s in place as of 1990 (baseline levels
include only basic or low enhanced I/M in some counties). There are then two additional
possible scenarios for further control:
We use two different regulatory scenarios in the paper:
69
Yes
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
1) Clean Air Act – This scenario represents the controls that regions have undertaken
or intend to undertake under the Clean Air mandates. States were asked what
programs they would implement to conform with CAA.
2) OTAG – This includes LEV in regions that indicated an intention to implement LEV.
Then there are all the possible least cost scenario results which depend on what is being
optimized. Under the least cost scenarios, regions are free to adopt whatever controls will be
lowest cost and still meet the regional nutrient or emissions reduction objectives. To do the
optimization, the counties are aggregated into about 74 groups. The groups are separated by
state, and within states by ozone attainment vs. non-attainment status, and MSA status (MSAs
are occasionally separate if they are large enough to be required to do I/M under the CAA100,000+ population, but are not part of a non-attainment area). In the Pechan dataset, LEV
vehicles could be introduced only by region: the possible regions are Massachusetts and New
York, the rest of the Ozone Transport Region (OTR) and states outside the OTR. In our least
cost analysis, LEV vehicles can be introduced in any of the 84 regions.
Table A3-2 shows the resulting controls under different scenarios.
Problems with the Mobile Source data:
There were several problems we found in the Pechan dataset on mobile sources used in this
analysis. We list them below and indicate what correction we took, if any, in each case.
1. The Pechan dataset assumed that I/M programs could only be applied to LDGVs
but in many regions light duty trucks are inspected as well (LDGT1). We
extended the possible set of options counties could implement to include I/M for
LDGT1.
2. Motorcycles were an 8th type of vehicle in the original dataset but were dropped
for this analysis because there were some inconsistencies in the data.
3. Emissions of VOC for LDGVs are higher under low enhanced I/M than they are
under basic I/M. This appears to be because of different assumptions in the
MOBILE Model which is used to forecast emission reductions under different I/M
scenarios. For example, basic I/M runs assumed 100% compliance whereas low
enhanced assumed 96% compliance. We have not changed these forecasts in this
version of the paper. However, it means that Low Enhanced I/M is unlikely to be
chosen in the least cost scenarios (higher cost and less emission reduction than
basic I/M).
4. Pechan revised its original cost estimate for high enhanced I&M from
$5.70/vehicle as listed in ERCAM-NOx, to $15.70/vehicle. We use the revised
estimate in this paper.
70
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
Table A3-2. Mobile Source NOx Abatement Technologies For CAA, OTAG
and Least Cost (OTAG) Scenarios
Mobile Source Technologies for Mobile Source Regions (n=84)
Abatement Options
Distribution of Technologies (%)
I&M
Reform
Gas
ReformDie
sel
LEV
CAA
OTAG
Least Cost
Scenario for
OTAG
0
0
0
0
0
0
2
2
2
2
2
4
4
4
4
4
4
6
6
6
6
6
6
6
0
0
0
1
1
1
0
0
1
1
1
0
0
0
1
1
1
0
0
0
0
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
6
8
0
6
8
0
8
0
6
8
0
4
6
0
4
6
0
5
7
9
0
5
7
38.1
0.0
0.0
3.6
0.0
0.0
3.6
0.0
4.8
0.0
0.0
20.2
0.0
0.0
4.8
0.0
0.0
7.1
0.0
0.0
0.0
17.9
0.0
0.0
0.0
26.2
11.9
0.0
1.2
2.4
0.0
3.6
0.0
1.2
3.6
0.0
4.8
15.5
0.0
2.4
2.4
0.0
0.0
1.2
6.0
0.0
2.4
15.5
27.4
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
3.6
9.5
56.0
3.6
0.0
0.0
0.0
Codes:
I&M: 0 = none; 2 = Basic; 4 = Low Enhanced; 6 = High Enhanced
Reformulated Gasoline: 0 = none; 1 = in place
Reformulated Diesel: 0 = none; 1 = in place
Low Emission Vehicles: 0 = none; 4 = Minimum National LEV (Mass. and NY); 5 = Maximum
National LEV (MA and NY); 6 = Minimum National LEV (other OTR states); 7 =
Maximum National LEV (other OTR states); 8 = Minimum National LEV (outside the
OTR); 9 = Maximum National LEV (outside the OTR)
71
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
5. The emission reductions assumed by LEVs in combination with High Enhanced
I/M are huge, and they come at a relatively low cost. Table A3-3 below shows the
emissions reductions and costs for various mobile source options in Allegheny
County. The emissions reductions are relative to the base case which is no mobile
source controls. It is clear from the table why either LEV with High Enhanced
I/M or High Enhanced I/M alone is chosen in the least cost scenario. We believe
these estimates are due to some very optimistic assumptions in the MOBILE
inventory Model (Harrington, McConnell and Cannon, 1998). We plan to
examine the sensitivity of our results to these assumptions, and to reassess some of
these estimates in future work.
Table A3-3. LDGV Emissions Reductions and Costs for Mobile Source Controls Allegany
County, Maryland
NOx Emissions
Reductions
Total Cost
Cost Effectiveness
(Tons/Ozone
($1990/Ozone
($1990/Ton NOx
Season)
Season)
Reduced)
No Controls
0
0
N/A
Reformulated Gasoline
27
261,189
9,811
Basic I&M
7
113,349
15,434
Low Enhanced I&M
8
113,349
14,817
High Enhanced I&M
110
312,208
2,846
LEV
21
92,866
4,398
LEV + Low Enhanced I&M
28
206,215
7,285
LEV + High Enhanced I&M
199
405,074
2,038
72
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
REFERENCES
Bloyd, Cary, et al. 1996. Tracking and Analysis Framework (TAF) Model Documentation
and User's Guide, ANL/DIS/TM-36, Argonne National Laboratory, December.
Chesapeake Executive Council. 1987. Chesapeake Bay Agreement, Annapolis, MD.
Clinton, President William J. 1997. Presidential Directive on the Implementation of New
National Ambient Air Quality Standards for Ozone and Fine Particulates.
Dennis, R. L. 1997. "Using the Regional Acid Deposition Model to Determine the Nitrogen
Deposition Airshed of the Chesapeake Bay Watershed," in Joel Baker, ed., Atmospheric
Deposition to the Great Lakes and Coastal Waters (Pensacola, FL: Society of
Environmental Toxicology and Chemistry), pp. 393-413.
Guthrie, Paul D., and Alan J. Krupnick. 1998. Air Quality Benefits Associated with
Reductions in Atmospheric Nitrogen Deposition to the Chesapeake Bay, Modeling
Methodology: Ozone and Fine Particles.
Guthrie, Paul D., Gerard E. Mansell, and Dongfen Gao. 1998. Assessment of Potential Air
Quality Benefits Associated with NOx Emissions reductions in the Chesapeake Bay
Airshed.
Harrington, Winston, V. McConnell, M. Cannon. 1998. A Behavioral Analysis of EPA's
MOBILE Emission Factor Model, Discussion Paper, Resources for the Future,
Washington, D.C., Summer.
Jones-Lee, M. W., M. Hammerton and P. R. Phillips. 1985. "The Value of Safety: Results
of a National Sample Survey," The Economic Journal, vol. 95, pp. 49-72.
Krupnick, Alan J., and Dallas Burtraw. 1996. "The Social Costs of Electricity: Do the
Numbers Add Up?" Resource and Energy Economics, vol. 18, pp. 423-466.
Krupnick, Alan J., Wallace E. Oates, and Eric Van den Berg. 1983. "On Marketable AirPollution Permits: The Case for a System of Pollution Offsets," Journal of Environmental
Economics and Management, vol. 10, no. 3 (September), pp. 233-247.
Lee, Russell., Alan Krupnick, Dallas Burtraw, and others. 1995. The External Costs and
Benefits of Fuel Cycles, A Study by the U.S. Department of Energy and the Commission
of the European Communities, prepared by Oak Ridge National Laboratory and
Resources for the Future (McGraw-Hill, Utility Data Institute, New Jersey), December.
Linker, L. C., G. E. Stigall, C. H. Chang, and A. S. Donigian. 1996. "Aquatic Accounting:
Chesapeake Bay Watershed Model Quantifies Nutrient Loads," Water Environment and
Technology, vol. 8, no. 1, pp. 48-52.
Miller, Paul, Praveen Amar, Marika Tatsutani. 1997. The Long-range Transport of Ozone
and its Precursors in the Eastern United States, NESCAUM, March.
Montgomery, David. 1972. "Markets in Licenses and Efficient Pollution Control Programs,"
Journal of Economic Theory, vol. 5, no. 3, December, pp. 395-418.
73
Krupnick, McConnell, Austin, Cannon, Stoessell, and Morton
RFF 98-46
Pechan, E. H., Associates, Inc. 1996a. Analysis of Costs and Benefits of a National Low
Emission Vehicle Program, EPA-68-D3-0035, Work Assignment No, II-67, prepared for
Vehicle Programs and Compliance Division, U.S. Environmental Protection Agency,
Washington, D.C., November.
Pechan, E. H., Associates, Inc. 1996b Atmospheric Nitrogen Deposition Loadings to the
Chesapeake Bay: An Initial Analysis of Cost Effectiveness of Control Options, EPA-68D3-0035, Work Assignment No. II-76, prepared for Chesapeake Bay Program Office,
U.S. Environmental Protection Agency, Annapolis, MD, November.
Pechan, E. H., Associates, Inc. 1997. The Emission Reduction and Cost Analysis Model for
NOx (ERCAM-NOx), EPA-68-D3-0035, Work Assignment No. II-63, prepared for Ozone
Policy and Strategies Group, U.S. Environmental Protection Agency, Research Triangle
Park, NC, September.
Rao, S. T., Tim Mount, and Gary Dorris. 1997. Least Cost Options for Ozone Improvement
in the Eastern United States: Estimating Emission Weights, prepared for the New York
State Energy Research and Development Authority.
Sisler, James F. 1996. Spatial and Seasonal Patterns and Long Term Variability of the
Composition of the Haze in the United States--An Analysis of Data from the IMPROVE
Network, Cooperative Institute for Research in the Atmosphere, Colorado State
University, July.
Tietenberg, Thomas H. 1985. Emissions Trading, an Exercise in Reforming Pollution Policy
(Washington, D.C.: Resources for the Future, Inc.).
U.S. Environmental Protection Agency (EPA). 1993. Regional Interim Emission Inventories
(1987-1991), Volume I: Development Methodologies. EPA-450/R-93-021a, U.S.
Environmental Protection Agency, Office of Air Quality Planning and Standards,
Research Triangle Park, NC, May.
U.S. Environmental Protection Agency (EPA). 1997. Finding of Significant Contribution
and Rulemaking for Certain States in the Ozone Transport Assessment Group Region for
Purposes of Reducing Regional Transport of Ozone. 62 FR 60318, U.S. Environmental
Protection Agency, November.
U.S. Environmental Protection Agency (EPA). 1998. Supplemental Notice for the Finding of
Significant Contribution and Rulemaking for Certain States in the Ozone Transport
Assessment Group Region for Purposes of Reducing Regional Transport of Ozone. 40
CFR Part 52, U.S. Environmental Protection Agency, March.
74